# Reachability Analysis: every page as markdown A reference that lists and does not rank. 37 headline CVEs disclosed from Apr to Nov 2026 with preconditions, fixes and mitigations, seven methods for deciding exposure, and a four-question tool that names the method that settles a finding. --- # Reachability analysis in four questions: which of your findings can be exploited REFERENCE, read 19 Nov 2026 **The short answer.** Reachability analysis checks whether vulnerable code is called and whether the service can be reached where you run it. The four-question tool on this page turns your answers into the method that settles a finding. The ledger covers 37 headline CVEs disclosed from 9 Apr to 5 Nov 2026, and 33 are in CISA KEV. ## Walk one finding through four questions Four questions, each answered yes, no or not sure: (1) Is the vulnerable code called? (2) Can the service be reached? (3) Is a control in front of it? (4) Is exploitation known? Start from your own finding or from one of the 37 ledger entries. Picking an entry fills the answers from its record: exploitation known from its KEV and first-exploitation dates, the question its settling method decides left as not sure, and the others set as exposed. An unanswered question counts as not sure; with all four open, Rule 4 sends the finding to Agent investigation. The tool opens on CVE-2026-123158 (OpenBSD OpenSSH, disclosed 5 Nov 2026, D+2), preselected in the entry list. Its four answers: (1) Is the vulnerable code called? not sure; (2) Can the service be reached? yes; (3) Is a control in front of it? no; (4) Is exploitation known? yes. Rule 5 applies: Reached: yes, no control in front, and the code question is the only one not sure. Method that settles it: [Static reachability and configuration reads](/methods/static-reachability/), or [runtime reachability and live-state reads](/methods/runtime-reachability/) on a running workload. The ledger records it for this entry because: Exposure rests on build and configuration facts that a read of the package and config settles: musl, no PAM, password authentication on, and a * account next to a hashed account. Those are static properties of the host, so a scan of the build and /etc/shadow gives the answer before any attack arrives, and a severity score or exploit feed cannot say which hosts meet all of them. - **What it proves:** That a call path exists from your code to the vulnerable function, shown step by step where one is found, and that the version, packaging and configuration facts in the CVE record hold or fail in the build. - **What it misses:** Network exposure, identity, controls, whether the code is deployed and running, and any setting changed after the read. - **Evidence to keep:** The call path from your code to the vulnerable function, with the path shown; or, from a running workload, whether the vulnerable package or function was loaded or executed, with the observation window. Static reachability reads code and build facts; runtime reachability reads a running workload and does not see code paths left unexercised during the window. - **Ledger:** 14 of 37 ledger entries settle on it, newest: [CVE-2026-123158](/ledger/cve-2026-123158/) (5 Nov 2026, D+2), [CVE-2026-19868](/ledger/cve-2026-19868/) (5 Nov 2026, D+1), [CVE-2026-76418](/ledger/cve-2026-76418/) (28 Oct 2026, D+0), [CVE-2026-111497](/ledger/cve-2026-111497/) (21 Oct 2026, D+5). ## Days from disclosure to first exploitation All 37 ledger entries by days from disclosure to first observed exploitation: 2 before disclosure, 10 on the day, 21 after, and 4 with no dated report. The median is 2 days. ## The ledger, all 37 entries | Disclosed | CVE | Product | Days | Code called | Reached | Control | Known | Settled by | |---|---|---|---|---|---|---|---|---| | 5 Nov 2026 | [CVE-2026-123158](/ledger/cve-2026-123158/) | OpenBSD OpenSSH | D+2 | open | yes | no | yes | Static reachability | | 5 Nov 2026 | [CVE-2026-19868](/ledger/cve-2026-19868/) | SonicWall SonicOS SSLVPN | D+1 | open | yes | no | yes | Static reachability | | 28 Oct 2026 | [CVE-2026-119977](/ledger/cve-2026-119977/) | WordPress Core | D+6 | open | yes | no | yes | Runtime reachability | | 28 Oct 2026 | [CVE-2026-76418](/ledger/cve-2026-76418/) | Cisco Secure Firewall ASA and FTD | D+0 | open | yes | no | yes | Static reachability | | 21 Oct 2026 | [CVE-2026-109618](/ledger/cve-2026-109618/) | PHP Group PHP | D+0 | yes | open | no | yes | Attack-path analysis | | 21 Oct 2026 | [CVE-2026-109619](/ledger/cve-2026-109619/) | PHP Group PHP | D+5 | yes | open | no | yes | Attack-path analysis | | 21 Oct 2026 | [CVE-2026-111497](/ledger/cve-2026-111497/) | Node.js Brotli decoder | D+5 | open | yes | no | yes | Static reachability | | 15 Oct 2026 | [CVE-2026-102764](/ledger/cve-2026-102764/) | BeyondTrust Remote Support | D-47 | open | yes | no | yes | Static reachability | | 14 Oct 2026 | [CVE-2026-0322](/ledger/cve-2026-0322/) | Palo Alto Networks PAN-OS | D+0 | yes | open | no | yes | Attack-path analysis | | 12 Oct 2026 | [CVE-2026-97229](/ledger/cve-2026-97229/) | GitLab CE/EE | D+1 | yes | open | no | yes | Attack-path analysis | | 5 Oct 2026 | [CVE-2026-21589](/ledger/cve-2026-21589/) | Atlassian Data Center products | D+10 | open | yes | no | yes | Static reachability | | 3 Oct 2026 | [CVE-2026-88779](/ledger/cve-2026-88779/) | Citrix NetScaler ADC and Gateway | D+1 | open | yes | no | yes | Static reachability | | 1 Oct 2026 | [CVE-2026-104286](/ledger/cve-2026-104286/) | Fortinet FortiMail | D+0 | open | yes | no | yes | Static reachability | | 30 Sep 2026 | [CVE-2026-94545](/ledger/cve-2026-94545/) | Vercel Satori and Next.js | no date | open | yes | no | no | Static reachability | | 22 Sep 2026 | [CVE-2026-94127](/ledger/cve-2026-94127/) | F5 BIG-IP APM | D+0 | open | yes | no | yes | Static reachability | | 16 Sep 2026 | [CVE-2026-76460](/ledger/cve-2026-76460/) | Cisco Identity Services Engine | D+0 | yes | open | no | yes | Attack-path analysis | | 8 Sep 2026 | [CVE-2026-81963](/ledger/cve-2026-81963/) | Microsoft Windows Update Stack | D+0 | yes | open | no | yes | Attack-path analysis | | 8 Sep 2026 | CVE-2026-85880 | Microsoft Windows ALPC | D+0 | yes | open | no | yes | Attack-path analysis | | 7 Sep 2026 | [CVE-2026-75650](/ledger/cve-2026-75650/) | Adobe Commerce and Magento Open Source | D-3 | yes | yes | no | yes | Known-exploited lists | | 4 Sep 2026 | [CVE-2026-87491](/ledger/cve-2026-87491/) | Google Chromium V8 | D+0 | open | open | open | yes | Agent investigation | | 3 Sep 2026 | [CVE-2026-84869](/ledger/cve-2026-84869/) | ConnectWise ScreenConnect | D+7 | open | open | open | yes | Agent investigation | | 28 Aug 2026 | [CVE-2026-82329](/ledger/cve-2026-82329/) | JFrog Artifactory | D+4 | open | open | open | yes | Agent investigation | | 26 Aug 2026 | [CVE-2026-80521](/ledger/cve-2026-80521/) | Linux kernel | no date | yes | open | no | no | Attack-path analysis | | 25 Aug 2026 | [CVE-2026-75604](/ledger/cve-2026-75604/) | Vercel Next.js | no date | open | yes | no | no | Static reachability | | 19 Aug 2026 | [CVE-2026-19490](/ledger/cve-2026-19490/) | Citrix NetScaler ADC and Gateway | D+21 | open | yes | no | yes | Static reachability | | 17 Aug 2026 | [CVE-2026-64849](/ledger/cve-2026-64849/) | MLflow Tracking Server | D+2 | open | yes | no | yes | Runtime reachability | | 12 Aug 2026 | [CVE-2026-42018](/ledger/cve-2026-42018/) | JFrog Artifactory | D+3 | yes | open | no | yes | Attack-path analysis | | 11 Aug 2026 | [CVE-2026-65660](/ledger/cve-2026-65660/) | Microsoft SharePoint Server | D+44 | open | yes | no | yes | Runtime reachability | | 11 Aug 2026 | [CVE-2026-68820](/ledger/cve-2026-68820/) | Microsoft Windows AFD for WinSock | D+0 | yes | open | no | yes | Attack-path analysis | | 29 Jul 2026 | [CVE-2026-41709](/ledger/cve-2026-41709/) | Broadcom VMware ESX | D+91 | yes | yes | no | yes | Known-exploited lists | | 27 Jul 2026 | [CVE-2026-42016](/ledger/cve-2026-42016/) | JFrog Artifactory | D+19 | yes | open | no | yes | Attack-path analysis | | 21 Jul 2026 | [CVE-2026-16723](/ledger/cve-2026-16723/) | Alibaba fastjson 1.x | no date | open | yes | no | no | Static reachability | | 8 Jul 2026 | [CVE-2026-59822](/ledger/cve-2026-59822/) | BerriAI LiteLLM | D+50 | open | yes | no | yes | Runtime reachability | | 2 Jul 2026 | [CVE-2026-9198](/ledger/cve-2026-9198/) | IBM Langflow OSS | D+15 | open | yes | no | yes | Runtime reachability | | 1 Jun 2026 | [CVE-2026-53266](/ledger/cve-2026-53266/) | Linux kernel | D+109 | open | yes | no | yes | Runtime reachability | | 13 May 2026 | [CVE-2026-46300](/ledger/cve-2026-46300/) | Linux Kernel | D+1 | open | yes | no | yes | Runtime reachability | | 9 Apr 2026 | [CVE-2026-34486](/ledger/cve-2026-34486/) | Apache Software Foundation Apache Tomcat | D+15 | open | yes | no | yes | Static reachability | Downloads: [ledger.csv](/datasets/ledger.csv), [ledger.json](/datasets/ledger.json), CC BY 4.0. ## The Methods Matrix at a glance A filled square means the method's proves text, as published on its page, speaks to the question; an empty one means it does not. An exploit-prediction score estimates how likely exploitation is across all deployments, so it speaks to the fourth question only in general, not for your environment. | Method | Code | Reach | Control | Known | | --- | --- | --- | --- | --- | | Agent investigation | proves it | proves it | proves it | no | | Attack-path analysis | no | proves it | proves it | no | | [EPSS exploit prediction](/methods/exploit-prediction/) | no | no | no | proves it | | [Known-exploited lists](/methods/known-exploited/) | no | no | no | proves it | | [Runtime reachability and live-state reads](/methods/runtime-reachability/) | proves it | no | no | no | | [Severity scores](/methods/severity-score/) | no | no | no | no | | [Static reachability and configuration reads](/methods/static-reachability/) | proves it | no | no | no | Alphabetical by method name. Each mark is read from the method's proves text. ## How the answers map to a method No standards body defines reachability analysis, so the primary source is vendor documentation. Snyk describes it as "identifying whether your application is calling a code element (for example, functions, classes, modules, and more) related to the vulnerability" ([Snyk reachability analysis documentation](https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis), read 19 Nov 2026). That answers the code question only. It gives the wrong answer when the call graph is incomplete, because the same page states that a NO PATH FOUND status does not mean the vulnerability is completely unreachable. It also says nothing about a server compromised before it was patched, as with CVE-2026-82329. The rules are read from the methods' own text and applied in order, and the first rule that matches applies. A "no" to reach, a "yes" to a control or a "no" to the code question rules exposure out, so the method is the one whose proves text produces that answer. Open questions go to the method that proves them; two or more open questions go to agent investigation, the one method whose proves text covers configuration, code, controls and network path. The fourth answer is read only when the finding is exposed, because known exploitation changes urgency, not exposure. The combinations column adds up to 81. | Rule | When the answers are | Method that settles it | Combinations of 81 | | --- | --- | --- | --- | | Rule 1 | Can the service be reached: no. The other answers do not change the method. | Attack-path analysis | 27 | | Rule 2 | Reached: yes or not sure, and a control is in front of it: yes. | Agent investigation | 18 | | Rule 3 | Reached: yes or not sure, no control confirmed, and the code is called: no. | [Static reachability and configuration reads](/methods/static-reachability/) | 12 | | Rule 4 | Nothing above rules it out, and two or three of the first three answers are not sure. | Agent investigation | 12 | | Rule 5 | Reached: yes, no control in front, and the code question is the only one not sure. | [Static reachability and configuration reads](/methods/static-reachability/), or [runtime reachability and live-state reads](/methods/runtime-reachability/) on a running workload | 3 | | Rule 6 | The code is called, and only the reach question or only the control question is not sure. | Attack-path analysis | 6 | | Rule 7 | Called: yes, reached: yes, control: no, and exploitation known: yes or not sure. | [Known-exploited lists](/methods/known-exploited/) | 2 | | Rule 8 | Called: yes, reached: yes, control: no, and exploitation known: no. | [EPSS exploit prediction](/methods/exploit-prediction/) | 1 | Three answers to four questions give 81 combinations; each one matches exactly one rule. Method text read 19 Nov 2026. ## Sources - Snyk reachability analysis documentation: definition of reachability analysis and the NO PATH FOUND limit: https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis (read 19 Nov 2026) - CVE-2026-123158: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-123158 (read 19 Nov 2026) - CVE-2026-123158: CISA KEV catalog entry, added 9 Nov 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-123158 (read 19 Nov 2026) - CVE-2026-123158: first exploitation, GreyNoise event page and timeline: https://sshole.events.greynoise.io/ (read 19 Nov 2026) - CVE-2026-19868: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-19868 (read 19 Nov 2026) - CVE-2026-19868: CISA KEV catalog entry, added 6 Nov 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-19868 (read 19 Nov 2026) - CVE-2026-119977: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-119977 (read 19 Nov 2026) - CVE-2026-119977: CISA KEV catalog entry, added 3 Nov 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-119977 (read 19 Nov 2026) - CVE-2026-76418: NVD record; the first-exploitation date, 28 Oct 2026, comes from this record: https://nvd.nist.gov/vuln/detail/CVE-2026-76418 (read 19 Nov 2026) - CVE-2026-76418: CISA KEV catalog entry, added 29 Oct 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-76418 (read 19 Nov 2026) - CVE-2026-109618: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-109618 (read 19 Nov 2026) - CVE-2026-109618: CISA KEV catalog entry, added 26 Oct 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-109618 (read 19 Nov 2026) - CVE-2026-109618: first exploitation, Fastly report on PHP2Shell exploitation, disclosure timeline and indicators: https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell (read 19 Nov 2026) - CVE-2026-109619: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-109619 (read 19 Nov 2026) - CVE-2026-109619: CISA KEV catalog entry, added 26 Oct 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-109619 (read 19 Nov 2026) - CVE-2026-111497: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-111497 (read 19 Nov 2026) - CVE-2026-111497: CISA KEV catalog entry, added 27 Oct 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-111497 (read 19 Nov 2026) - CVE-2026-111497: first exploitation, ELLIO research on exploitation and affected builds: https://ellio.tech/en/blog/node2shell-cve-2026-111497-rce-on-stock-nodejs-builds/ (read 19 Nov 2026) - CVE-2026-102764: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-102764 (read 19 Nov 2026) - CVE-2026-102764: CISA KEV catalog entry, added 16 Oct 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-102764 (read 19 Nov 2026) - CVE-2026-102764: first exploitation, BeyondTrust blog on ARMORY CASTLE and BT26-05: https://www.beyondtrust.com/blog/entry/bt26-05 (read 19 Nov 2026) - CVE-2026-0322: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-0322 (read 19 Nov 2026) - CVE-2026-0322: CISA KEV catalog entry, added 14 Oct 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-0322 (read 19 Nov 2026) - CVE-2026-97229: NVD record; the first-exploitation date, 13 Oct 2026, comes from this record: https://nvd.nist.gov/vuln/detail/CVE-2026-97229 (read 19 Nov 2026) - CVE-2026-97229: CISA KEV catalog entry, added 29 Oct 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-97229 (read 19 Nov 2026) - CVE-2026-21589: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-21589 (read 19 Nov 2026) - CVE-2026-21589: CISA KEV catalog entry, added 15 Oct 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-21589 (read 19 Nov 2026) - CVE-2026-88779: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-88779 (read 19 Nov 2026) - CVE-2026-88779: CISA KEV catalog entry, added 4 Oct 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-88779 (read 19 Nov 2026) - CVE-2026-104286: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-104286 (read 19 Nov 2026) - CVE-2026-104286: CISA KEV catalog entry, added 1 Oct 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-104286 (read 19 Nov 2026) - CVE-2026-94545: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-94545 (read 19 Nov 2026) - CVE-2026-94127: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-94127 (read 19 Nov 2026) - CVE-2026-94127: CISA KEV catalog entry, added 22 Sep 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-94127 (read 19 Nov 2026) - CVE-2026-76460: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-76460 (read 19 Nov 2026) - CVE-2026-76460: CISA KEV catalog entry, added 16 Sep 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-76460 (read 19 Nov 2026) - CVE-2026-81963: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-81963 (read 19 Nov 2026) - CVE-2026-81963: CISA KEV catalog entry, added 8 Sep 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-81963 (read 19 Nov 2026) - CVE-2026-85880: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-85880 (read 19 Nov 2026) - CVE-2026-85880: CISA KEV catalog entry, added 8 Sep 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-85880 (read 19 Nov 2026) - CVE-2026-75650: NVD record carrying Adobe's advisory data for APSB26-146 (affected versions, hotfix, CVSS): https://nvd.nist.gov/vuln/detail/CVE-2026-75650 (read 19 Nov 2026) - CVE-2026-75650: CISA KEV catalog entry, added 8 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-75650 (read 19 Nov 2026) - CVE-2026-75650: first exploitation, Sansec research: StyleSmuggler 0-day (hotfix details, versions tested, indicators, timeline): https://sansec.io/research/stylesmuggler-0day (read 19 Nov 2026) - CVE-2026-87491: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-87491 (read 19 Nov 2026) - CVE-2026-87491: CISA KEV catalog entry, added 9 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-87491 (read 19 Nov 2026) - CVE-2026-87491: first exploitation, Microsoft Edge security release notes: https://learn.microsoft.com/en-us/deployedge/microsoft-edge-relnotes-security (read 19 Nov 2026) - CVE-2026-84869: NVD record; the first-exploitation date, 10 Sep 2026, comes from this record: https://nvd.nist.gov/vuln/detail/CVE-2026-84869 (read 19 Nov 2026) - CVE-2026-84869: CISA KEV catalog entry, added 11 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-84869 (read 19 Nov 2026) - CVE-2026-82329: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-82329 (read 19 Nov 2026) - CVE-2026-82329: Wiz Research report, which confirmed successful exploitation between 1 and 8 Sep 2026, so servers were compromised before they were patched; also the first-exploitation source for CVE-2026-42018, 15 Aug 2026; also the first-exploitation source for CVE-2026-42016, 15 Aug 2026: https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201 (read 19 Nov 2026) - CVE-2026-82329: CISA KEV catalog entry, added 2 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-82329 (read 19 Nov 2026) - CVE-2026-82329: first exploitation, Fastly Security Research report, 3 Sep 2026: https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity (read 19 Nov 2026) - CVE-2026-80521: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-80521 (read 19 Nov 2026) - CVE-2026-75604: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-75604 (read 19 Nov 2026) - CVE-2026-19490: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-19490 (read 19 Nov 2026) - CVE-2026-19490: CISA KEV catalog entry, added 9 Sep 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-19490 (read 19 Nov 2026) - CVE-2026-64849: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-64849 (read 19 Nov 2026) - CVE-2026-64849: CISA KEV catalog entry, added 19 Aug 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-64849 (read 19 Nov 2026) - CVE-2026-42018: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-42018 (read 19 Nov 2026) - CVE-2026-42018: CISA KEV catalog entry, added 11 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42018 (read 19 Nov 2026) - CVE-2026-65660: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-65660 (read 19 Nov 2026) - CVE-2026-65660: CISA KEV catalog entry, added 25 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-65660 (read 19 Nov 2026) - CVE-2026-65660: first exploitation, Previdian research on exploitation attempts: https://blog.previdian.com/cve-2026-65660-previdian-observes-two-stage-sharepoint-exploitation-attempts/ (read 19 Nov 2026) - CVE-2026-68820: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-68820 (read 19 Nov 2026) - CVE-2026-68820: CISA KEV catalog entry, added 11 Aug 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-68820 (read 19 Nov 2026) - CVE-2026-41709: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-41709 (read 19 Nov 2026) - CVE-2026-41709: CISA KEV catalog entry, added 28 Oct 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-41709 (read 19 Nov 2026) - CVE-2026-42016: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-42016 (read 19 Nov 2026) - CVE-2026-42016: CISA KEV catalog entry, added 11 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42016 (read 19 Nov 2026) - CVE-2026-16723: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-16723 (read 19 Nov 2026) - CVE-2026-59822: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-59822 (read 19 Nov 2026) - CVE-2026-59822: CISA KEV catalog entry, added 2 Sep 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-59822 (read 19 Nov 2026) - CVE-2026-59822: first exploitation, Wiz Research honeypot report: https://www.wiz.io/blog/ai-infrastructure-honeypot (read 19 Nov 2026) - CVE-2026-9198: NVD record; the first-exploitation date, 17 Jul 2026, comes from this record: https://nvd.nist.gov/vuln/detail/CVE-2026-9198 (read 19 Nov 2026) - CVE-2026-9198: CISA KEV catalog entry, added 4 Aug 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-9198 (read 19 Nov 2026) - CVE-2026-53266: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-53266 (read 19 Nov 2026) - CVE-2026-53266: CISA KEV catalog entry, added 18 Sep 2026; the first-exploitation date is this KEV date added: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-53266 (read 19 Nov 2026) - CVE-2026-46300: NVD record; the first-exploitation date, 14 May 2026, comes from this record: https://nvd.nist.gov/vuln/detail/CVE-2026-46300 (read 19 Nov 2026) - CVE-2026-46300: CISA KEV catalog entry, added 28 Oct 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-46300 (read 19 Nov 2026) - CVE-2026-34486: NVD record: https://nvd.nist.gov/vuln/detail/CVE-2026-34486 (read 19 Nov 2026) - CVE-2026-34486: CISA KEV catalog entry, added 4 Aug 2026: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-34486 (read 19 Nov 2026) - CVE-2026-34486: first exploitation, SOCRadar research on the SNOWLIGHT campaign (exploitation and indicators): https://socradar.io/blog/snowlight-government-chinese-campaign/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ --- # A vulnerable system holds a flaw; an exploitable one also meets every condition an attacker needs Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** A system is vulnerable when it holds a weakness that, in NIST's words, "could be exploited." It is exploitable when every precondition in the advisory also holds on it: the configuration, the network path, the privilege. Exploited is a third fact, about the world, shown by evidence such as the CISA KEV listing of 9 Nov 2026. ## Three words describe three different facts NIST's glossary defines a vulnerability as a "Weakness in an information system, system security procedures, internal controls, or implementation that could be exploited or triggered by a threat source" (FIPS 200, from CNSSI 4009). The definition says "could". It describes what a flaw permits in principle. It says nothing about whether this system, as configured today, lets anyone do it. CISA's Known Exploited Vulnerabilities (KEV) catalog uses a different test. A CVE enters the catalog when "There is reliable evidence that the vulnerability has been actively exploited in the wild" (alongside an assigned CVE ID and a clear remediation action). That is a statement about attackers somewhere, not about your host. Exploitable sits between the two. It is the claim that every precondition the advisory names is true on one specific system. Each state has its own question and its own evidence. | State | Question | Answered by | Who usually holds the fact | | --- | --- | --- | --- | | Vulnerable | Does the affected code exist here? | Installed version or build, compared with the advisory's ranges | Scanner, package manager, SBOM | | Exploitable here | Does every precondition hold on this system? | Configuration, enabled features, network path, accounts, the running process | System owner, network team | | Exploited | Is there reliable evidence of use against anyone? | A KEV listing, a vendor statement, incident reports | CISA, the vendor, threat intelligence | ## A worked case: CVE-2026-123158 on six SSH hosts OpenSSH 10.6p2 was released on 5 Nov 2026. NVD published CVE-2026-123158 the same day: sshd "on Linux, when musl libc is used without PAM, lacks a workaround for a password authentication bypass", and for some account configurations "any password between 256 and 1024 characters succeeds". CISA added it to KEV on 9 Nov 2026, four days later. So on 9 Nov the exploited fact is settled for every host. The other two facts are not. Alpine and the researcher give the length as 257 to 1024 bytes where NVD says 256 to 1024 characters; this page uses Alpine's figure, because it is the vendor advisory for the hosts below. The Alpine Linux advisory of 5 Nov 2026 names Alpine 3.16 through 3.24 and edge as affected. Fixed openssh-server packages by branch: edge 10.6_p2-r0, 3.24 10.3_p1-r2, 3.23 10.2_p1-r1, 3.22 10.0_p1-r11, 3.21 9.9_p2-r1, 3.20 9.7_p1-r6, 3.19 9.6_p1-r3, 3.18 9.3_p2-r4, 3.17 9.1_p1-r7 and 3.16 9.0_p1-r6. The advisory lists what must hold together on a host. Each condition has a command that answers it, and an owner who runs it and keeps the output. Run the commands as root where /etc/shadow is read: - Platform team, musl libc, not glibc: ldd /usr/sbin/sshd | grep -i musl - Platform team, no PAM in use: Alpine's openssh-server package is built without PAM; openssh-server-pam with UsePAM no counts as no PAM. Run sshd -T | grep -i usepam on the PAM package. - System owner for the setting, network team for the path: sshd -T | grep -i passwordauthentication must print yes, and the SSH port must be reachable from where an attacker can stand. - System owner, an account whose password field is "*" (root on Alpine base images, key-only users, the alpine user in tiny-cloud images): `awk -F: '$2=="*" {print $1}' /etc/shadow` - System owner, at least one local account with a $5$ or $6$ crypt hash: `awk -F: '$2 ~ /^\$[56]\$/ {n++} END {print n+0}' /etc/shadow` - Platform team, the running sshd is the fixed build. Alpine states that sshd must be restarted after the package upgrade, because upgrading alone does not replace the running daemon. ## Five of six hosts are flagged and two can be attacked The estate below was built for this page to show the spread. A scanner that compares the openssh package version with the range "before 10.6p2" flags five of the six hosts. Two are exploitable. One of those two is not flagged at all. Host lab-5 is the case a version count gets backwards. Its package was upgraded to the fixed build, so the scanner is quiet, yet sshd was never restarted. The old daemon still answers on port 22 with password login on and the account conditions met. | Host | Setup | Version-only scanner | Precondition that decided it | State on 19 Nov 2026 | | --- | --- | --- | --- | --- | | bastion-1 | Alpine 3.22, openssh-server before 10.0_p1-r11, password login on, port 22 open to the internet, root:* and a deploy account with a $6$ hash | Flags | None fail | Exploitable, and exploited elsewhere (KEV 9 Nov 2026) | | build-2 | Same image as bastion-1 | Flags | PasswordAuthentication is no | Vulnerable, not exploitable | | app-3 | Ubuntu, glibc, PAM in use, OpenSSH version number below 10.6p2 | Flags | Not musl, PAM in use; NVD limits the flaw to musl without PAM | Not affected, a scanner false positive | | edge-4 | Alpine 3.23, openssh-server-pam with UsePAM yes | Flags | UsePAM is yes | Not affected | | lab-5 | Alpine 3.22, package upgraded to 10.0_p1-r11 or later, sshd not restarted, same accounts and settings as bastion-1 | Quiet | Running process is still the old build | Exploitable | | ci-6 | Alpine container image with openssh-server installed, sshd never started | Flags | No sshd process | Not exposed | ## Exploited can attach to a low CVSS score or a narrow precondition KEV listing and exploitability are separate, and they do not move together. CVE-2026-41709 in VMware ESX is rated Low by Broadcom, with a CVSS v3 base score of 2.7. Broadcom's advisory VMSA-2026-0006 says "A malicious administrator could exploit this issue to perform certain operations without them being logged." The advisory was first published on 29 Jul 2026. CISA listed the CVE on 28 Oct 2026, 91 days later. The precondition, an attacker who is already an ESX administrator, did not change. The reason to act did. CVE-2026-0322 in PAN-OS was listed in KEV on its publication day, 14 Oct 2026. Palo Alto Networks describes an "authenticated attacker with network access" to the management interface, and says the risk "is minimized by restricting access to the management interface to only trusted internal IP addresses". A firewall whose management plane is closed to untrusted networks is vulnerable, and is exploited elsewhere, but is not exploitable from outside. CVE-2026-76460 in Cisco ISE sits at the other end. Cisco says it affects ISE and ISE-PIC "regardless of device configuration", so the only conditions are the release and a reachable API. For a finding like that, vulnerable and exploitable are close to the same set. ## A version range can be wider or narrower than the exposure NVD describes CVE-2026-75604 in Next.js as affecting applications "using Pages Router or App Router without Cache Components on Windows-hosted servers", from 13.4.0 until the fixes in 15.5.24 and 16.3.3. A Next.js 15.2 application on a Linux container is inside the version range and outside the condition. It is not in the KEV feed read on 19 Nov 2026. The reverse also happens, as the lab-5 row shows: a version check says patched while the live process is not. In both directions the version is evidence about files, and the advisory's preconditions are about a running system. ## Record the state, the deciding fact and the date The four-question tool on the home page asks the same things in a fixed order: is the code called, can the service be reached, is a control in front of it, is exploitation known. Starting it from the CVE-2026-123158 ledger entry sets exploitation known to yes, which is true of every host here. The per-host answers to the other three are the facts in the table above, and a not-sure on reach or control sends the finding to the method the tool names. A decision can be reviewed and reversed only if it names its fact. The record for bastion-1 and build-2 on 19 Nov 2026 has these fields: - Finding: CVE-2026-123158, OpenSSH, bastion-1 and build-2. - State: bastion-1 exploitable; build-2 vulnerable, not exploitable. - Deciding fact: build-2 sshd -T prints passwordauthentication no. bastion-1 prints yes, with root:* and one $6$ hash present. - Evidence: command output saved with the host name and time; KEV listing dated 9 Nov 2026. - Action and due date: patch bastion-1 and restart sshd now; build-2 follows the normal patch cycle. - Recheck when: the advisory is revised, or PasswordAuthentication changes on build-2. ## Related pages - [How to prove a finding is not exploitable](/guides/prove-a-finding-is-not-exploitable/) - [How agent investigation records its evidence](/guides/how-agent-investigation-records-its-evidence/) - [Reachability analysis in practice](/guides/reachability-analysis-in-practice/) ## Sources - NIST CSRC glossary: vulnerability: https://csrc.nist.gov/glossary/term/vulnerability (read 19 Nov 2026) - CISA: Reducing the significant risk of known exploited vulnerabilities (KEV criteria): https://www.cisa.gov/known-exploited-vulnerabilities (read 19 Nov 2026) - CISA KEV catalog JSON feed (release 2026.11.18): https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - NVD record for CVE-2026-123158: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-123158 (read 19 Nov 2026) - OpenSSH release notes, 10.6p2: https://www.openssh.org/releasenotes.html (read 19 Nov 2026) - Alpine Linux advisory, OpenSSH password authentication bypass, 5 Nov 2026: https://alpinelinux.org/posts/2026-11-05-openssh-password-authentication-bypass/ (read 19 Nov 2026) - Broadcom advisory VMSA-2026-0006 (CVE-2026-41709): https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017 (read 19 Nov 2026) - Palo Alto Networks advisory for CVE-2026-0322: https://security.paloaltonetworks.com/CVE-2026-0322 (read 19 Nov 2026) - Cisco advisory cisco-sa-ISE-ABP-VNSW7Tn5 (CVE-2026-76460): https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5 (read 19 Nov 2026) - NVD record for CVE-2026-75604: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/exploitable-vs-vulnerable/ --- # Prove a finding is not exploitable by naming the failed precondition and keeping dated evidence for it Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** Proving a finding is not exploitable means showing that one precondition in the advisory fails on your system, with evidence a second person can repeat. Worked case: a fastjson 1.2.76 service in a Tomcat WAR, closed against CVE-2026-16723 on 19 Nov 2026 with four dated checks and a review date. ## A closure is a failed precondition, not a feeling of safety A finding is not exploitable on a system when one condition the advisory requires is false there. The advisory must name the condition, and your evidence must show it is false on this system, on a date. The first condition that fails ends the test. The record holds that one condition, the proof the advisory still lists it, and nothing else. A [CVSS score](/methods/severity-score/) cannot stand in for that proof. The FIRST CVSS v4.0 user guide says Base scores "are designed to measure the severity of a vulnerability and should not be used alone to assess risk". ## Five kinds of evidence can fail a precondition Version and build evidence is mechanical and the strongest. Configuration and deployment evidence changes with the system, so each needs a date and a re-check trigger. The examples are preconditions from the ledger records. | Evidence | Ledger precondition | What makes the claim stale | | --- | --- | --- | | Version outside the affected range | CVE-2026-88779 Citrix NetScaler ADC and Gateway: builds before 14.1-73.41 and 13.1-64.28 are affected, and those builds and later are fixed | Restoring an older build from backup or downgrading | | Deployment form excludes the trigger | CVE-2026-16723 fastjson: the advisory says plain jar, uber-jar and Tomcat or Jetty WAR deployments do not meet the trigger | Repackaging as a Spring Boot executable fat-jar | | Feature or role not enabled | CVE-2026-94127 BIG-IP APM: affected only as an OAuth Authorization Server | A profile change | | Platform excludes it | CVE-2026-75604 Next.js: Linux and macOS hosts are not described as affected | Moving the workload to a Windows host | | Interface not reachable | CVE-2026-0322: needs network access to the management interface | A firewall or routing change | ## Worked case: fastjson 1.2.76 in a Tomcat WAR, closed against CVE-2026-16723 The finding. On 19 Nov 2026 a scanner flags invoice-api, a service used here as an illustration, for fastjson 1.2.76. CVE-2026-16723 is a remote code execution flaw in fastjson 1.2.68 to 1.2.83. The advisory, dated 21 Jul 2026 and updated 29 Jul 2026, rates it Critical and lists the conditions: the version range, AutoType and SafeMode off (the defaults), and "Spring Boot executable fat-jar (i.e., launched via java -jar xxx.jar)". Its not-affected table lists "Non-fat-jar deployments", including "Tomcat/Jetty WAR". The ledger records no KEV entry for it. The test starts with the cheapest condition to falsify, which is the deployment form. Version 1.2.76 is inside the range, so the version cannot close the finding, and the other two conditions are properties of defaults. The steps below are what the engineer runs. The output shown is the illustration's output. - E1, version. In the build checkout, run `mvn dependency:tree -Dincludes=com.alibaba:fastjson`. Output: `com.alibaba:fastjson:jar:1.2.76:compile`. This confirms the finding is real and in range. - E2, artifact shape. Run `unzip -l invoice-api-4.8.2.war | grep -c BOOT-INF` and read the manifest with `unzip -p invoice-api-4.8.2.war META-INF/MANIFEST.MF`. Output: 0 BOOT-INF entries, and a manifest with no Main-Class line. A Spring Boot executable archive carries a launcher entry point, and this one has none. - E3, how it runs. On host web-12 run `ps -o args= -C java` and `ls /opt/tomcat/webapps`. Output: one process whose arguments start `org.apache.catalina.startup.Bootstrap`, and `invoice-api.war` unpacked under webapps. No process runs `java -jar` on any invoice-api archive. - E4, pipeline. Read the deploy job definition in the CI system. It copies the WAR into /opt/tomcat/webapps and restarts Tomcat. It has no step that starts a jar. The owner of the pipeline confirms in the ticket. ## The record that closes the finding The record cites E1 to E4 by label and holds no copy of their output, which is attached to the ticket. A second person, not the engineer who ran the checks, signs it after reading the attachments and the advisory page. | Field | Entry | | --- | --- | | Claim | Not exploitable on invoice-api, host web-12, as of 19 Nov 2026 | | Advisory and revision | CVE-2026-16723, fastjson2 wiki advisory dated 21 Jul 2026, updated 29 Jul 2026 | | Precondition tested | Deployed as a Spring Boot executable fat-jar launched with java -jar | | Result | False: the artifact is a WAR with no Spring Boot launcher, run by Tomcat (E2, E3, E4) | | Finding still real | Yes: fastjson 1.2.76 is present (E1), so the dependency stays on the upgrade list | | Run by | Service engineer, 19 Nov 2026, build 4.8.2 | | Signed off by | Security reviewer, 19 Nov 2026, after repeating E2 | | Stale if | The artifact is rebuilt as an executable jar or executable WAR run with java -jar, the advisory is revised, or CVE-2026-16723 is added to KEV | | Review date | 19 Feb 2027, or at the first trigger | ## A closure holds only while the stale conditions stay false Wire each stale condition to an event. A change to the build packaging should reopen the record. So should an advisory revision, or a KEV listing of the CVE. A calendar date alone misses all three. The claim also stays narrow: 'not exploitable on host web-12 as of 19 Nov 2026' ages better than 'not exploitable'. The closure leaves fixing the dependency a separate decision. The advisory's fixes are fastjson 1.2.84, SafeMode, a noneautotype build or a move to fastjson2, and they remove the dependence on the deployment form. ## Some advisories leave no precondition to fail The ledger records Cisco's position for CVE-2026-76460: ISE and ISE-PIC on a release without the patch are affected with no configuration dependency, so a configuration argument has nothing to test. Atlassian says instances with user authentication should still be treated as exposed for CVE-2026-21589, so only the fixed version or restricted network access changes the answer. Two more cases need care. GitLab says CVE-2026-97229 is exploitable "under certain conditions" and does not publish them, so the honest record reads 'not provable'. The fastjson advisory itself leaves one gap: it does not say how a Spring Boot executable WAR run with java -jar is treated. Evidence for that packaging must show the manifest and the launch command, as E2 and E3 do above. Timing can also flip a result. WordPress CVE-2026-119977 is reachable only during a full-package core upgrade, and sites with automatic background updates open that window without an administrator acting. A check run on a quiet day proves nothing about the upgrade window. ## Related pages - [Reachability analysis in practice](/guides/reachability-analysis-in-practice/) - [Exploitable vs vulnerable](/guides/exploitable-vs-vulnerable/) - [Vulnerability triage with evidence](/guides/vulnerability-triage-with-evidence/) ## Sources - fastjson advisory for CVE-2026-16723, dated 21 Jul 2026, updated 29 Jul 2026: https://github.com/alibaba/fastjson2/wiki/Security-Advisory:-Remote-Code-Execution-in-fastjson-1.2.68%E2%80%931.2.83 (read 19 Nov 2026) - FIRST CVSS v4.0 user guide: https://www.first.org/cvss/v4.0/user-guide (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/prove-a-finding-is-not-exploitable/ --- # Reachability analysis runs in two passes: a code pass for a path, a runtime pass for a call Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** Reachability analysis checks whether the vulnerable code element is called by your software, from the code (static) or from a running workload (runtime). The worked run below takes CVE-2026-111497, a Brotli decoder flaw in Node.js, through both passes on 19 Nov 2026: the application code had no Brotli reference, a dependency did, and a probe saw the call. ## The method asks whether the vulnerable element is used Snyk describes the method as "identifying whether your application is calling a code element (for example, functions, classes, modules, and more) related to the vulnerability", and reports one of three statuses: REACHABLE ("A direct or indirect path was found from your application to the vulnerable code"), NO PATH FOUND, and NOT APPLICABLE ("Reachability is not supported for this vulnerability"). Semgrep Supply Chain pairs the affected dependency versions with "the pattern for the vulnerable code" and scans the codebase for it. Endor Labs builds static call graphs over the project and its dependencies, which it says often comprise over 100 packages. The runtime variant watches the workload instead. Endor Labs describes instrumented container reachability as embedding "a runtime sensor in your container image to record how your application uses the image in a real environment". The two passes need different inputs and miss different things. Static analysis needs source or build artifacts and lockfiles. Its limit, in Snyk's wording, is that "just because there is no evidence of this does not mean that the element cannot be reached": reflection and dynamic behavior can hide a call. Runtime analysis needs a sensor or probe on a live process and an observation window. Its limit is the window: code that does not run during the window leaves no trace. Neither pass tells you whether an attacker can reach the input, which is a separate exposure question. ## Step 1 is to name the vulnerable element from the advisory CVE-2026-111497 is a flaw in the Brotli decoder. NVD lists Google Brotli 1.0.3 through 1.2.0, fixed in 1.0.10, 1.1.1 and 1.2.1. The Node.js security release of 21 Oct 2026 bundles the fix in v22.23.4, v24.22.1 and v26.12.1 and says an attacker "who can supply Brotli data that an application decompresses with node:zlib can corrupt process memory". CISA listed it in KEV on 27 Oct 2026. That sentence names the element: a Brotli decompression stream created through node:zlib. The reachability question for any service on an affected Node.js line is whether anything in the process creates one. Whoever owns the service runs it, and the record it leaves is the answer, the evidence and the date. ## Step 2, the static pass: the application is clean and a dependency is not The sample service, report-api, is an 8-line Node.js program written for this page. It calls a partner feed with axios 1.21.1 and returns the length of the body. Its source never mentions Brotli or node:zlib. A search of the application code is the first static check: grep -rniE "brotli|node:zlib|require('zlib')" app.js printed nothing. Read alone, that is a NO PATH FOUND result, and the usual answer is wrong here. The advisory says "an application, or a library it calls", so the search has to cover the dependency tree: grep -rlE "createBrotliDecompress|brotliDecompress" node_modules printed one file, node_modules/axios/lib/adapters/http.js. Reading it shows two relevant lines. Line 57 builds the request header as 'gzip, compress, deflate' plus ', br' when zlib.createBrotliDecompress exists. Line 1328 calls zlib.createBrotliDecompress when the response header says content-encoding br. The path is app.js, axios.get, the http adapter, zlib.createBrotliDecompress. In Snyk's vocabulary that is REACHABLE, with a call path. A text search is a crude call graph: it finds the call site but not whether the branch can execute, and it would miss a call made through a computed name. Tools in this category build the graph across the whole dependency tree for that reason. ## Step 3, the runtime pass: a probe records the call and the window decides the result The probe is a file loaded before the application with node --require ./probe.js app.js. It replaces zlib.createBrotliDecompress with a wrapper that prints the caller each time a decoder is created. The first version assigned the wrapper with a plain assignment, zlib.createBrotliDecompress = function ... . It ran silently and recorded zero calls, because the assignment had no effect on the zlib export. That is a false NOT LOADED from a bad probe, the failure to watch for in any runtime sensor. Replacing the assignment with Object.defineProperty fixed it. Window one: the partner upstream replied with content-encoding br, after seeing axios send Accept-Encoding: gzip, compress, deflate, br. The probe printed: probe: createBrotliDecompress call 1 from at RedirectableRequest.handleResponse (/tmp/demo/node_modules/axios/dist/node/axios.cjs:4951:33). The request returned 100 bytes. Window two: the same service, the same request, but the upstream ignored the br offer and sent the body uncompressed. The probe printed nothing. Whether the decoder runs is decided by what the other side sends, so a quiet window proves little. A runtime result is true for the window that was watched. ## Step 4 is to record each answer with its limit The machine used for this run reported Node.js v22.23.5, at or above the fix in v22.23.4, so this machine is not affected. The table is what the record would say for an image of the same service on a 22.x build before v22.23.4. | Question | Method | Answer | Evidence | Limit | | --- | --- | --- | --- | --- | | Is the affected build present? | Version check | Depends on the image: below v22.23.4 is affected | node --version in each image | Says nothing about use | | Does the application code call Brotli? | Static, application code | No hits | grep of app.js, 19 Nov 2026 | Misses libraries | | Does a dependency call it? | Static, dependency tree | Yes, axios http adapter, lines 57 and 1328 | grep of node_modules plus reading the file, 19 Nov 2026 | Call site found, branch not proven | | Was a decoder created at runtime? | Runtime probe | Yes in window one, no in window two | Probe output with call stack, 19 Nov 2026 | True for the window watched | | Can an attacker control the compressed bytes? | Exposure check, outside both passes | Open: depends on who controls the partner feed and the network path | Not tested in this run | Needs its own evidence | ## Where the usual answer is wrong A clean search of your own code does not clear a finding when the advisory names a library path, as the axios result shows. A status of NO PATH FOUND is, in Snyk's words, not a statement that the vulnerability "is completely unreachable or unexploitable", and Snyk notes that results labeled this way "can change to Reachable over time" as the analysis improves. Re-run after a dependency update. Some findings are not reachability questions at all. CVE-2026-75604 in Next.js turns on the host being Windows and the router type, and CVE-2026-123158 in OpenSSH turns on musl, PAM and account state. A call graph has nothing to say about either. Read the advisory's preconditions first and choose the method after. ## Documented offerings, alphabetically The vendors below each document at least one of the two passes. - Endor Labs: static call graphs across project and dependencies; instrumented container reachability with a runtime sensor. - Semgrep: Supply Chain rules that pair affected versions with a code pattern. - Snyk: call-graph analysis of application against dependencies, with REACHABLE, NO PATH FOUND and NOT APPLICABLE statuses. ## Related pages - [Vulnerability triage with evidence](/guides/vulnerability-triage-with-evidence/) - [How to prove a finding is not exploitable](/guides/prove-a-finding-is-not-exploitable/) - [Reading a vendor advisory's preconditions](/guides/reading-a-vendor-advisory/) ## Sources - Snyk reachability analysis: https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis (read 19 Nov 2026) - Semgrep Supply Chain overview: https://semgrep.dev/docs/semgrep-supply-chain/overview (read 19 Nov 2026) - Endor Labs call graphs: https://docs.endorlabs.com/scan/sca/call-graphs/index (read 19 Nov 2026) - Endor Labs instrumented container reachability: https://docs.endorlabs.com/scan/containers/instrumented-reachability/index (read 19 Nov 2026) - NVD record for CVE-2026-111497: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-111497 (read 19 Nov 2026) - Node.js security releases, 21 Oct 2026: https://nodejs.org/en/blog/vulnerability/october-2026-security-releases (read 19 Nov 2026) - CISA KEV catalog JSON feed (release 2026.11.18): https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - axios 1.21.1 package, lib/adapters/http.js, from the npm registry: https://registry.npmjs.org/axios/1.21.1 (read 19 Nov 2026) - NVD record for CVE-2026-75604: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604 (read 19 Nov 2026) - NVD record for CVE-2026-123158: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-123158 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/reachability-analysis-in-practice/ --- # Triage with evidence asks the cheap questions first and stores every answer Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** Triage with evidence asks 5 questions in order and stores each answer: is the affected version installed, does each advisory precondition hold, is the asset reachable from where an attacker can stand, is the bug exploited, and what is the fix and who changes the asset. Cheap questions come first, so most findings close early. ## The order follows cost and time A backlog is a list of version matches. Treating each match as equal spends the week on findings that cannot be exploited here while the few that can wait in the queue. Timing is the reason to hurry. In the 37-CVE ledger on this site, read on 19 Nov 2026, 33 CVEs show dated exploitation. Of those, 24 were exploited within seven days of disclosure (3 of them before it), and the median is 2 days after disclosure. CISA's BOD 26-04 states the aim as "efficiently prioritizing high-risk vulnerabilities for timely action, while deferring action against low-risk vulnerabilities"; triage here means that sorting. The directive sets remediation urgency from four variables: "Asset Exposure", "KEV Status", "Exploit Automation" and "Technical Impact". Steps 3 and 4 below answer the first two. CISA publishes the other two for each CVE through its Vulnrichment program, and step 4 stores them beside the KEV answer. Two cheaper checks sit in front: whether the affected build is installed at all, and whether the advisory's preconditions hold. The four-question tool on the home page takes the answers to steps 2 to 4 (code called, service reached, control in front of it, exploitation known) and names the method that settles any open one. ## Five steps, who answers each and where each fails Stop at the first step that fails and write down which one. A step that cannot be answered is recorded as unknown with a next action, never as a pass. | Step | Question | Source of the answer | Who answers | Usually fails when | | --- | --- | --- | --- | --- | | 1 | Is the affected version or build installed? | Inventory, lockfile, package list, appliance version | Inventory owner | The inventory shows the product but not the hotfix level | | 2 | Does each advisory precondition hold? | Vendor advisory read against the configuration | Analyst with configuration access | The advisory withholds a condition, or the configuration is read from memory | | 3 | Is it reachable from where an attacker can stand? | Network path, management-plane exposure, runtime state | Network or cloud engineer | Exposure is assumed from the diagram instead of tested | | 4 | Is it exploited, and can it be automated? | CISA KEV listing or a dated vendor or research report; Automatable and Technical Impact where CISA or the vendor publishes them | Analyst, by lookup | A vendor says "exploitation observed" with no date | | 5 | What is the fix, and who changes the asset? | Advisory fixed versions, ownership data | Asset owner | The fixed build for the running release train is not copied exactly | ## One advisory carried through four Cisco ISE nodes The advisory is Cisco cisco-sa-ISE-ABP-VNSW7Tn5, CVE-2026-76460: an unauthenticated API request to Cisco ISE or ISE-PIC bypasses the web management interface, and Cisco warns of command execution with root privileges. Cisco published it on 16 Sep 2026 with the statement that its PSIRT is aware of active exploitation. CISA added it to KEV the same day with a due date of 19 Sep 2026 and forensic triage required. CVSS 3.1 base score 10.0. The nodes below are invented for the example; the patch levels, dates and quotations are real. Opened 16 Sep 2026, four nodes in scope. Node ise-a runs 3.4 with Patch 7 installed, the first fixed patch for the 3.4 train, so it closes at step 1. Node ise-b is a pilot on 3.0, which is end of software maintenance: Cisco gives it no fix and advises moving to a supported release, so it goes to step 5 as a migration. Nodes ise-c (3.2 Patch 8) and ise-d (3.5 Patch 2) are below their trains' fixed patches, 3.2 Patch 11 and 3.5 Patch 4. Step 2 for ise-c and ise-d: Cisco says the flaw affects ISE and ISE-PIC "regardless of device configuration", so no setting can close it and the only open condition is whether the API can be reached. Step 4 for both: KEV listed, due date 19 Sep 2026, three days after listing. CISA's SSVC data in NVD gives automatable yes and technical impact total, the two other BOD 26-04 variables. The nodes differ at step 3. ise-c sits behind an infrastructure access control list that admits only the admin VLAN, which is the mitigation Cisco describes. That lowers who can reach the API and does not fix the flaw, so the finding stays open. ise-d serves a guest portal on a network that guests and unmanaged devices can reach. KEV marks forensic triage as required for the CVE, so it applies to both nodes, and BOD 26-04 defines it as remediation or mitigation within three days plus a check whether the system is compromised. The ACL narrows who could reach ise-c but does not show that no admin-VLAN host sent the request, and ise-c was unpatched while exploitation was active. The network team applies the ACL to ise-d on 16 Sep 2026, the security team starts the compromise check on both nodes, and both are upgraded by the due date: 3.2 Patch 11 for ise-c and 3.5 Patch 4 for ise-d. Outcome: four findings, one closed at step 1 with a reason, one migration ticket, two upgrade tickets with a named owner and the exact target patch, each with a compromise check. CISA's due dates bind US federal agencies; for everyone else they are a benchmark. ## The record to store for each finding Tickets built from this record carry their reasoning, so the assignee does not repeat the analysis. Every field has a date, because a stale answer is the usual reason a closed finding is wrong. - Finding: CVE ID, asset name, date opened, analyst. - Build: the version found, the command or export it came from, the date it was read. - Advisory: URL, the revision date shown on it, and the sentence quoted for each precondition. - Preconditions: one line each marked holds, fails or unknown, with the evidence and its date. - Reachability: from which networks the interface answers, and how that was tested. - Exploitation: KEV added and due dates, or the vendor's statement quoted with its date, or the words "none stated"; Automatable and Technical Impact where published. - Fix: the exact target build for this release train, the owner, the change ticket and the due date. - Decision: the step reached and the closing reason, or the ticket link. - Reopen triggers: advisory revision, a KEV change, or a change to the asset's build or exposure. ## When exploitation came first, the patch does not close the finding BeyondTrust Remote Support CVE-2026-102764 runs the order backwards. The actor exploited a zero-day on 29 Aug 2026, 47 days before the advisory BT26-05 of 15 Oct 2026, and CISA listed it on 16 Oct 2026 with forensic triage required. For a self-hosted instance that was in scope and unpatched in that period, step 4 comes first and a compromise check is part of closing. If compromise is found, rotate every credential the appliance stored or injected. SaaS instances were patched by the vendor, which closes them at step 1. ## Unknown is not a pass, and the count of unknowns shows where to invest For a KEV-listed bug on an internet-facing system, treat an unknown at step 3 as reachable until shown otherwise, because being wrong costs far more than checking. For a local-only bug with no listing, an unknown can wait for the next patch cycle. CVE-2026-80521, a Linux AF_UNIX garbage collector use-after-free, is an example: CVSS AV:L and PR:L mean the attacker needs unprivileged code execution first, and it is absent from KEV release 2026.11.18, with an [EPSS](/methods/exploit-prediction/) of 0.0018. Count unknowns by step each month. A pile at step 3 means missing network and runtime visibility. A pile at step 5 means missing ownership data. ## Where this guide stops This guide covers the order and the record. Proving that a precondition fails, with the evidence that survives a second reader, is covered in how to prove a finding is not exploitable. Deciding whether code can call a vulnerable function is covered in reachability analysis in practice. Dependency findings have their own guide for step 1 to 3 detail. ## Related pages - [Reading a vendor advisory's preconditions](/guides/reading-a-vendor-advisory/) - [Reachability analysis in practice](/guides/reachability-analysis-in-practice/) - [Triaging a dependency finding](/guides/triaging-a-dependency-finding/) ## Sources - Cisco advisory cisco-sa-ISE-ABP-VNSW7Tn5 (CVE-2026-76460): https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5 (read 19 Nov 2026) - CISA KEV JSON feed, release 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - CISA BOD 26-04: Prioritizing Security Updates Based on Risk: https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk (read 19 Nov 2026) - BeyondTrust BT26-05 blog and timeline: https://www.beyondtrust.com/blog/entry/bt26-05 (read 19 Nov 2026) - BeyondTrust advisory BT26-05: https://www.beyondtrust.com/trust-center/security-advisories/bt26-05 (read 19 Nov 2026) - NVD record for CVE-2026-102764: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-102764 (read 19 Nov 2026) - NVD record for CVE-2026-76460: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-76460 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/vulnerability-triage-with-evidence/ --- # Read an advisory phrase by phrase: the conditions sit in the description, the exposure note and the workaround Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** Vendor advisories hide the conditions for exploitation in short phrases. This guide reads Palo Alto Networks advisory CVE-2026-0322 field by field, from the summary block to the solution table, then maps stock phrases from Atlassian, BeyondTrust, Cisco, Fortinet and GitLab advisories to the action each one calls for. ## The summary block is the vendor's opinion, so decode each field The advisory for CVE-2026-0322 opens with a summary block before any prose. The vendor supplies every field. The [CVSS](/methods/severity-score/) 4.0 specification defines the supplemental ones, and the vector at the foot of the page carries them as short codes. | Field and value | Vector code | What it says | What it leaves open | | --- | --- | --- | --- | | Severity 8.6 HIGH | AV:N/AC:L/AT:N/PR:H/UI:N | Network attack, low complexity, no extra requirements, high privileges needed | Whether anyone outside holds an administrator account | | Exploit Maturity ATTACKED | E:A | Exploitation has been seen | When it began and how many victims | | Automatable NO | AU:N | Attackers cannot reliably automate all four kill chain steps across targets | Whether a determined attacker can still do it by hand | | Recovery USER | R:U | A person must act to restore service after an attack | Whether a root-level compromise can be undone by restoring service | | Value Density DIFFUSE | V:D | The vendor judges the system holds limited resources; the specification's example of concentrated is a central email server | Your view of a perimeter firewall, which is yours to make | | Response Effort MODERATE | RE:M | The vendor expects a moderate effort to respond | Your change window and outage cost | | Urgency HIGHEST | U:Red | Provider Urgency, an optional supplemental metric in which the provider assesses "the highest urgency" (Red) | Your own priority: the value is the provider's pass-through assessment, and your exposure, accounts and change window still set your action | ## The description holds two conditions, a scope sentence and one mitigation The description says an "authenticated attacker with network access" can run arbitrary code "with root privileges". That is two conditions, a network path to the management interface and a valid account, plus the impact. Read "root" as full device control, not as a service-level compromise. The scope sentence is the one that decides step 1: the issue applies to PAN-OS on PA-Series and VM-Series firewalls and on Panorama, virtual and M-Series. "Cloud NGFW and Prisma Access are not impacted." A finding raised on a hosted service closes here. The sentence that follows, saying risk is minimized by restricting management access to trusted internal IP addresses, is a mitigation placed in the description. It is easy to skip, and it is the only workaround-like text outside the workarounds section. ## "Required Configuration for Exposure" says none, then adds a second reading The section opens with "No special configuration is required to be affected by this issue." Default builds are in scope, so a clean configuration review cannot close the finding. The next sentence says the risk is highest when management access is allowed from external IP addresses on the internet. The advisory prices the difference. Restricting access to a single jump box changes the vector from AV:N to AV:A and lowers the CVSS-B and CVSS-BT, both 8.6, to 8.5. The vendor therefore treats a jump box as a small reduction, not as removal of the exposure. ## The product status table and the solution table cut versions differently Product Status lists affected ranges per sub-train, such as "< 11.2.10-h15". The Solution table lists the running range and the build to move to. A firewall on 11.2.10-h9 sits in "11.2.8 through 11.2.10-h*", where the "h*" covers any hotfix, and the answer is "Upgrade to 11.2.10-h15 or 11.2.16 or later." Moving it to 11.2.7-h21, which is the target for a lower range, would be a downgrade. The last row, "All older unsupported PAN-OS versions", says "Upgrade to a supported fixed version." It means a firewall on an end-of-life release is affected and has no patch on its own train. ## "Very limited exploitation has been observed" gives no date The Exploitation Status section is one sentence. Together with Exploit Maturity ATTACKED it supports step 4 of a triage record, but it names no start date. CISA's KEV entry, dated 14 Oct 2026 and the same day as publication, is the dated fact to cite. The page also says the issue was discovered internally and shows an update on 15 Oct 2026, the date to store as the revision you read. "Very limited" describes what the vendor has seen. It is not a statement about what is possible against your build. ## "No known workarounds exist" is followed by a conditional signature The workarounds section says no workaround exists, then offers Threat IDs 510046 and 510047 from content version 9159-10287. They work only if management traffic is routed through a data-plane port, the certificate is replaced, inbound traffic is decrypted and threat prevention is enabled. The FAQ explains "limited coverage": variations may exist that the signature cannot detect, and the vendor recommends upgrading. Record such signatures as detection, not as a closing reason. The FAQ also lists three IP addresses tied to exploit attempts. They are useful for log searches, and an empty result does not show the device is clean. ## The same phrases in other vendors' advisories Stock phrases recur across vendors. Read the phrase, then read what surrounds it. | Phrase and source | What it says | What to do | | --- | --- | --- | | "regardless of device configuration" (Cisco, CVE-2026-76460) | No configuration removes exposure | Patch, and restrict access to the nodes; the advisory's mitigation is infrastructure access control lists | | "There are no workarounds that address this vulnerability. However, there is a mitigation." (Cisco) | A control reduces remote exploitation but is temporary | Record it as mitigation with a date, and keep the upgrade ticket open | | "The Cisco PSIRT is aware of active exploitation" (Cisco) | Exploitation exists, no date given | Cite the advisory revision, then check KEV for a date | | "under certain conditions" (GitLab, CVE-2026-97229) | A condition exists and is not stated; the phrase appears on several other entries in the same release too, so it carries no signal | Treat reachable self-managed instances as exposed and patch to 19.4.2, 19.3.5 or 19.2.8 | | "Under the right conditions" (BeyondTrust BT26-05, CVE-2026-102764) | A condition exists and details are withheld | Apply the patch; check for earlier compromise | | "Affected Atlassian Cloud products have been patched" and "No Cloud customer action is required" (Atlassian, CVE-2026-21589) | The vendor patched the hosted service | Scope work to Data Center, Crucible and Fisheye | | "Exploitation requires prior knowledge of the target file's exact name and path" (Atlassian) | A real precondition that limits impact | Record it, and still patch: the advisory says some configurations hold sensitive files | | "reported to be exploited in the wild, customers are urged to apply the workaround below" (Fortinet FG-IR-26-175, CVE-2026-104286) | Exploited, and the workaround is the first action | Disable the IBE service or limit webmail access today, then upgrade | | "Upgrade to upcoming 7.2.10 or above" and "Migrate to a fixed release" (Fortinet) | No fixed 7.2 build yet; 7.0 gets none | Apply the workaround and track the date | | "need not perform any action" for FortiMail Cloud, remediated 5 Oct 2026 (Fortinet) | Hosted service fixed after the 1 Oct 2026 publication; the timeline shows a clarification on 7 Oct 2026 | Store the remediation date from the advisory | ## Fixed-version text needs your release line, not the headline number Fortinet lists a separate fix for each FortiMail line: 8.0.2, 7.6.7 and 7.4.9. BeyondTrust ties the patch to a version range: the BT26-05 patch covers 24.3.1 to 26.2.3, and installs older than 24.3.1 must upgrade first. Copy the line that matches the running build, and record the exact build after patching, since "patched" without a version cannot be checked later. A record line for the CVE-2026-0322 advisory, as stored on 19 Nov 2026: CVE-2026-0322, PAN-OS, advisory updated 15 Oct 2026; CVSS-B 8.6, vector CVSS:4.0/AV:N/AC:L/AT:N/PR:H/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N/E:A/AU:N/R:U/V:D/RE:M/U:Red; conditions: network path to the management interface and a high-privilege account; scope: PA-Series, VM-Series, Panorama, not Cloud NGFW or Prisma Access; exploitation: ATTACKED, no start date, KEV 14 Oct 2026; fix: the build for the running sub-train, for example 11.2.10-h15 or 11.2.16 on 11.2.8 to 11.2.10-h*; workaround: none, Threat IDs 510046 and 510047 as detection only. ## Related pages - [Triaging a dependency finding](/guides/triaging-a-dependency-finding/) - [Vulnerability triage with evidence](/guides/vulnerability-triage-with-evidence/) - [Triaging an internet-facing appliance finding](/guides/triaging-an-internet-facing-appliance-finding/) ## Sources - Palo Alto Networks advisory for CVE-2026-0322: https://security.paloaltonetworks.com/CVE-2026-0322 (read 19 Nov 2026) - FIRST CVSS v4.0 specification document: https://www.first.org/cvss/v4.0/specification-document (read 19 Nov 2026) - CISA KEV JSON feed, release 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - Cisco advisory for CVE-2026-76460: https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5 (read 19 Nov 2026) - GitLab patch release 19.4.2: https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-4-2-released/ (read 19 Nov 2026) - BeyondTrust advisory BT26-05: https://www.beyondtrust.com/trust-center/security-advisories/bt26-05 (read 19 Nov 2026) - Atlassian advisory for CVE-2026-21589: https://confluence.atlassian.com/security/cve-2026-21589-arbitrary-file-access-vulnerability-impacts-multiple-products-1870495748.html (read 19 Nov 2026) - Fortinet advisory FG-IR-26-175: https://fortiguard.fortinet.com/psirt/FG-IR-26-175 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/reading-a-vendor-advisory/ --- # A dependency finding is settled by the resolved version, the packaging and a call site, each checked with a command Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** Ask the build tool for the resolved version, compare it with the advisory's affected and not-affected rows, inspect how the artifact is packaged, then search for a call site fed by untrusted data. For fastjson CVE-2026-16723, the string 1.2.83 gave not affected as a noneautotype build and affected as a plain pin; 1.2.84 is the fix. ## The resolved version comes from the build tool, not the manifest Maven's documentation defines the term: "Maven avoids the need to discover and specify the libraries that your own dependencies require by including transitive dependencies automatically." A manifest states what the team asked for. The build tool states what shipped, and a transitive copy only shows up there. In Maven, mvn dependency:tree -Dincludes=com.alibaba prints just the matching branches. In npm, npm ls followed by the package name does the same. For a library bundled inside a runtime, such as Brotli in Node.js, the check is node --version on each image. The service team runs the command in CI or on the built artifact. The record is the command, its output and the date, pasted into the ticket. The step fails when someone reads the manifest instead, or runs the command on one module of a multi-module build and misses a copy that arrives through another. ## Worked pass one: fastjson CVE-2026-16723 in a Spring Boot service The fastjson advisory, dated 21 Jul 2026 and updated 29 Jul 2026, lists 1.2.68 to 1.2.83 as affected and 1.2.84 as the fix. The trigger needs AutoType off, SafeMode off and a Spring Boot executable fat-jar. CVE-2026-16723 is not in the CISA KEV feed dated 18 Nov 2026, so the urgency comes from the [CVSS](/methods/severity-score/) 3.1 base score of 9.0 and the reachability answer, not from known exploitation; its [EPSS](/methods/exploit-prediction/) of 0.0066 is low. To get real tool output we built a Spring Boot 3.3.5 service on 19 Nov 2026 whose only fastjson arrives through com.alibaba.csp:sentinel-transport-common 1.8.6, then ran mvn dependency:tree -Dincludes=com.alibaba. It printed: [INFO] demo:orders-api:jar:1.0.0 [INFO] \- com.alibaba.csp:sentinel-transport-common:jar:1.8.6:compile [INFO] \- com.alibaba:fastjson:jar:1.2.83_noneautotype:compile The version reads 1.2.83, inside the affected range. The artifact is the noneautotype build, and the advisory's not-affected table has a row for it: "noneautotype builds", reason "Vulnerable code removed at compile time". A scanner that trims the suffix would flag this service. The tree output settles it, and the ticket records the artifact name. Then we pinned com.alibaba:fastjson to 1.2.83 in dependencyManagement and ran the same command: the tree showed com.alibaba:fastjson:jar:1.2.83:compile. Pinning to 1.2.84 showed 1.2.84. The three outcomes: | Resolved artifact | What the advisory says | Outcome | Action and owner | | --- | --- | --- | --- | | fastjson 1.2.83_noneautotype, through sentinel-transport-common 1.8.6 | noneautotype builds: vulnerable code removed at compile time | Not affected by CVE-2026-16723 | Service team records the tree output and the advisory row, then closes | | fastjson 1.2.83, pinned in dependencyManagement | 1.2.68 to 1.2.83 affected when the app is a fat-jar with AutoType and SafeMode off | Affected if a call site parses untrusted JSON | Service team upgrades to 1.2.84, or sets -Dfastjson.parser.safeMode=true until the release ships | | fastjson 1.2.84 | Fixed: type names with URL-special characters are rejected | Not affected | Close with the tree output | ## The packaging check decides whether the trigger can happen at all The advisory names a deployment form, so read it from the built artifact. On the test service with fastjson pinned to 1.2.83, unzip -p target/orders-api-1.0.0.jar META-INF/MANIFEST.MF printed Main-Class: org.springframework.boot.loader.launch.JarLauncher, and unzip -l showed BOOT-INF/lib/fastjson-1.2.83.jar nested inside. That is the executable fat-jar shape. The advisory's not-affected table lists "plain java -jar, uber-jar, Tomcat/Jetty WAR" as not meeting the trigger condition, so the same code built as a WAR for Tomcat falls in that row. The platform team usually knows the packaging from the pipeline. Record the manifest line, because packaging changes when someone switches build plugins. The step fails when the check is run on the source tree or a local build instead of the artifact that is deployed. ## A call site that parses untrusted JSON makes the finding real Search the service source with grep -rnE "JSON\.(parse|parseObject)\(" src/main and classify each hit by where the string comes from. Request bodies, queue messages and upstream responses count. A parse of a file bundled in the jar does not. The advisory warns that "Specifying a target class (e.g., JSON.parseObject(body, SomeDto.class) ) is not a mitigation", so a typed call still counts when the class has Object or Map fields. The code owner writes the list of file and line with a conclusion for each. A static reachability tool can build the same list across many services where it supports the language, and reflection can hide calls from it, so a clean result is evidence, not proof. The step fails in both directions: reflection hides a real call, and a hit on a test class or a bundled file raises a finding that cannot be triggered. The four-question tool on the home page records this search as its first question, is the code called. When that is the only open answer and the service is reached with no control in front, the tool names static reachability, or runtime reachability on a running workload. ## Worked pass two: Brotli CVE-2026-111497 in a Node.js service The Node.js security release of 21 Oct 2026 covers lines 26.x, 24.x and 22.x and says an attacker who can supply Brotli data that an application decompresses with node:zlib can corrupt process memory. Fixed releases are v22.23.4, v24.22.1 and v26.12.1. CISA added the CVE to KEV on 27 Oct 2026, six days after the release, so this one is a known-exploited finding. On a v22.23.5 runtime, node --version printed v22.23.5, later than the fixed v22.23.4. Compare the Node.js release number with the advisory's list and record that. The embedded string node -p "process.versions.brotli" is not a test: ELLIO reports it shows 1.2.0 on patched and vulnerable Node.js builds alike, because the fix was cherry-picked without changing the version string. For a runtime below the fixed release, the next question is the call path. We installed got 11.8.6 and ran npm ls decompress-response, which printed got@11.8.6 and decompress-response@6.0.0. In the installed source, got sets accept-encoding to gzip, deflate, br when process.versions.brotli is a string, and decompress-response calls zlib.createBrotliDecompress() when the reply says content-encoding br. Any service that uses got to call an upstream therefore decompresses Brotli from that upstream, and the attacker condition becomes whether someone can control or alter that upstream reply. The decision follows from that. If the runtime is below the fixed release and a Brotli path exists, the ticket names the fixed release for the line as the target and cites the KEV date. Until the upgrade ships, the vendor mitigation is to avoid negotiating or decompressing br from untrusted peers. If no path exists, record the grep and the npm ls output, and still schedule the upgrade, since the vulnerable decoder remains in the runtime. ## Four cases where the version string gives the wrong answer Two are above: the fastjson artifact suffix and the Node.js Brotli string. A third is that an older release can be unaffected: the fastjson advisory says versions up to 1.2.60 lack the vulnerable code path. A fourth is that a short affected list can still apply to you. The Apache Tomcat security pages list CVE-2026-34486 as affecting only 11.0.20, 10.1.53 and 9.0.116, the three releases that carried a flawed fix for CVE-2026-29146, and it matters only when a cluster uses EncryptInterceptor. Apache fixed it in 11.0.21, 10.1.54 and 9.0.117. The checks are catalina.sh version and grep -n EncryptInterceptor conf/server.xml on each node. CISA listed the CVE on 4 Aug 2026. A package unused in one service is still resolved by the others, so run the commands per service. ## The ticket carries the evidence, the decision and the owner Fields to fill: the resolved version and the command that showed it; the packaging line from the manifest; the call sites with a conclusion each; the decision (upgrade, interim control, or not affected with the advisory row cited); the target version; the owner and the date. A reader who gets that ticket six months later can see which fact decided the outcome. The step fails when the ticket says "not affected" without the advisory row or the command output behind it. ## Related pages - [Triaging an internet-facing appliance finding](/guides/triaging-an-internet-facing-appliance-finding/) - [Reading a vendor advisory's preconditions](/guides/reading-a-vendor-advisory/) - [How agent investigation records its evidence](/guides/how-agent-investigation-records-its-evidence/) ## Sources - fastjson security advisory for CVE-2026-16723: https://github.com/alibaba/fastjson2/wiki/Security-Advisory:-Remote-Code-Execution-in-fastjson-1.2.68%E2%80%931.2.83 (read 19 Nov 2026) - NVD record for CVE-2026-16723: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-16723 (read 19 Nov 2026) - Maven Central listing for com.alibaba:fastjson: https://repo1.maven.org/maven2/com/alibaba/fastjson/ (read 19 Nov 2026) - Node.js October 2026 security releases: https://nodejs.org/en/blog/vulnerability/october-2026-security-releases (read 19 Nov 2026) - NVD record for CVE-2026-111497: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-111497 (read 19 Nov 2026) - CISA KEV entry for CVE-2026-111497: https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-111497 (read 19 Nov 2026) - npm registry entry for got: https://registry.npmjs.org/got (read 19 Nov 2026) - NVD record for CVE-2026-34486 (Apache Tomcat): https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-34486 (read 19 Nov 2026) - CISA KEV feed, dated 18 Nov 2026: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - Apache Maven: Introduction to the dependency mechanism: https://maven.apache.org/guides/introduction/introduction-to-dependency-mechanism.html (read 19 Nov 2026) - ELLIO on CVE-2026-111497 (process.versions.brotli and patched builds): https://ellio.tech/en/blog/node2shell-cve-2026-111497-rce-on-stock-nodejs-builds/ (read 19 Nov 2026) - Apache Tomcat 11 security page (CVE-2026-34486, fixed in 11.0.21; the 10 and 9 pages give 10.1.54 and 9.0.117): https://tomcat.apache.org/security-11.html (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/triaging-a-dependency-finding/ --- # For an internet-facing appliance, check the build, the enabled feature and the exposure, then check for compromise even after patching Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** 4 checks settle an appliance finding: read the build from the device, compare it with the vendor's first fixed release in your line, confirm the feature the advisory names is on and reachable, then search for the vendor's indicators. FortiMail, PAN-OS and Cisco ASA/FTD flaws reached CISA's KEV list within a day of disclosure. ## Read the build from the device, not from the asset list An appliance bundles the operating system, the service and the configuration, so it is patched as one unit and rarely takes a scanner agent. The network or platform team reads the version on the device: get system status on FortiMail, show system info | match sw-version on PAN-OS, and show version on a Cisco ASA or FTD. Paste the output into the ticket with the date. Then find your release line in the vendor's table. Take the first fixed release for that line. A later release in the same line is fine, and an earlier one is not. PAN-OS fixes are per maintenance branch: a firewall on 11.2.9 sits in the advisory's 11.2.8 up to, not including, 11.2.10-h15 row, so its fix is 11.2.10-h15 or 11.2.16. The step fails when the version is copied from an asset list that was last updated before an upgrade, or when a hotfix suffix is dropped. The table holds the two lines used in the examples here. Every release line is in the ledger entries for [CVE-2026-104286](/ledger/cve-2026-104286/), [CVE-2026-0322](/ledger/cve-2026-0322/) and [CVE-2026-76418](/ledger/cve-2026-76418/). | Product and CVE | Release line | First fixed release | | --- | --- | --- | | FortiMail, CVE-2026-104286 | 7.4 (7.4.0 to 7.4.8 affected) | 7.4.9 | | PAN-OS, CVE-2026-0322 | 11.2.8 up to, not including, 11.2.10-h15 | 11.2.10-h15 or 11.2.16 or later | ## Confirm the feature is on and the interface is reachable The security team checks the configuration the advisory names, because it decides whether a vulnerable build is exposed. The first fixed release stays the target either way. The interim control differs by product. The step fails when the check is run on the standby unit of a pair only, or when the setting is read from a template instead of the running device. - FortiMail: the flaw is unauthenticated and in the GUI component. Check Encryption, IBE, IBE Service in the GUI and whether webmail is reachable from the internet. The advisory does not state IBE as a condition, but its first workaround turns the service off and its log lines name the IBE decrypter. The workaround is config system encryption ibe, set status disable, end. Alternatives are limiting webmail to trusted private networks, or blocking POST requests to /ibe that contain ../ at a web application firewall. - PAN-OS: Palo Alto Networks states that no special configuration is required to be affected, and "The risk is highest when you allow access to the management interface from external IP addresses on the internet." The attacker also needs a valid high-privilege account on the management interface, so read Device > Administrators and the authentication profiles as well as the permitted IP addresses. The vendor lists no workaround; the [advisory guide](/guides/reading-a-vendor-advisory/) decodes the jump-box and Threat ID text. - Cisco ASA and FTD: the device is exposed when webvpn is enabled on an interface and a tunnel group uses secondary authentication. Run show running-config webvpn and show running-config tunnel-group, and look for secondary-authentication-server-group in every group, including DefaultWEBVPNGroup. Cisco notes "This check is not limited to profiles that are displayed in the group list on the login page." Cisco lists no workaround. Its mitigation is no tunnel-group-list enable when the default group has no secondary authentication, or no secondary-authentication-server-group when it does, and Cisco recommends the upgrade instead. ## Patching does not show whether the device was already compromised CISA's BOD 26-04 defines the step: the agency must "carry out a forensic triage of the asset to assess whether the system is compromised". A device that was reachable while the flaw was known to attackers needs that check after the patch, and BOD 26-04 asks "Is the vulnerable asset publicly exposed?" to decide how soon. The security team runs it, searching the full log retention and not only the days since disclosure, and records the search terms, the period and the result. The pace is short: CVE-2026-0322 in PAN-OS was published and listed in KEV on 14 Oct 2026, and Cisco published CVE-2026-76418 on 28 Oct 2026 and CISA listed it on 29 Oct 2026. Each ledger entry holds the vendor's indicators for its product: - FortiMail: the [CVE-2026-104286](/ledger/cve-2026-104286/) entry lists the source addresses, the system event and encryption log lines, and the archive-account check. - PAN-OS: the [CVE-2026-0322](/ledger/cve-2026-0322/) entry lists the addresses tied to exploit attempts. An empty search does not show the device is clean. - Cisco ASA and FTD: the [CVE-2026-76418](/ledger/cve-2026-76418/) entry notes that the advisory publishes no indicators. The effect is a device reload, so compare the uptime line of show version and the reload history with change records. - BeyondTrust Remote Support: the [CVE-2026-102764](/ledger/cve-2026-102764/) entry holds the BT26-05 timeline. BeyondTrust says the actor rotates infrastructure for every victim, so hunt behavior. If compromise is found, rotate every credential the appliance stored or injected, review new multi-factor registrations on privileged accounts, and image the appliance before rebuilding it. ## A worked case: a FortiMail 7.4.6 gateway, 1 to 7 Oct 2026 Take a gateway running 7.4.6 with webmail published to external recipients. The gateway and its search results are an illustration; the advisory facts are real. Fortinet published FG-IR-26-175 on 1 Oct 2026 as Critical, CVSS v3 9.8, Attack Type Unauthenticated, Known Exploited Yes. CISA added CVE-2026-104286 to KEV the same day, with a due date of 4 Oct 2026 and forensic triage required. 1 Oct: get system status shows 7.4.6, inside 7.4.0 through 7.4.8, and webmail faces the internet. The mail administrator reads Encryption > IBE and finds the service on. Fortinet does not state IBE as a condition of the flaw; the ledger entry infers it from the workaround and the log lines. With an in-range build, an unauthenticated flaw and an exposed interface, the device is treated as affected whatever the IBE reading, and exploitation is already stated. The administrator turns IBE off with the CLI above the same day and saves the configuration before and after. 4 Oct, the CISA due date: the security team searches the system event and encryption logs from the start of retention, 1 Sep 2026, and the firewall and proxy logs for the two indicator addresses. Result: 0 matches for the two indicator addresses in the ledger entry, no archive account the mail team did not create, no cron line running /migadmin, no Invalid Base64 lines. The ticket records the period 1 Sep to 4 Oct 2026, the terms and the result, with the security team as owner. A hit would mean the box is treated as compromised: preserve the logs, rotate administrator and archive credentials, and rebuild from a known-good image. The upgrade to 7.4.9 follows in the booked window. 5 Oct: Fortinet updates the solution table and remediates FortiMail Cloud. For this on-premises box the target is 7.4.9. The change window is booked with the mail team as owner. 7 Oct: the advisory update clarifies that cloud customers need no action. That does not apply to this device. ## When the usual answer is wrong A configuration check clears a device only when the advisory names a configuration. Palo Alto states that no special configuration is required for CVE-2026-0322, so no setting read can rule out a PAN-OS firewall on an affected build: only the fixed release or a restricted management address changes the answer. A high availability pair is two devices. The standby holds its own build and its own logs, so read the version on both units, and search the logs of both, because a patched active unit paired with an unpatched standby is exposed after a failover. Turning off the feature, as the FortiMail workaround does, closes the flaw from that day. It says nothing about the days before, so the compromise search still covers the period from the start of log retention. ## The ticket names the build, the fix and the check owner Write the answer in the title: product, build in service, first fixed release for the line, exploitation status. For example: FortiMail 7.4.6 gateway, CVE-2026-104286, exploited, upgrade to 7.4.9, compromise search due by a stated date. The body lists the version output, the feature check, the interim control with its date, the indicator search with its period and result, and a named owner for each. Some fixes need an outage window, so the infrastructure owner books it when the ticket opens. The four-question tool on the home page fits before the ticket: for a flaw on the management plane or webmail, set service reached and control in front from the checks above, and the tool names the method that settles the open answer. Its rule list sends a closed-off management interface to attack-path analysis. ## Related pages - [How agent investigation records its evidence](/guides/how-agent-investigation-records-its-evidence/) - [Triaging a dependency finding](/guides/triaging-a-dependency-finding/) - [Exploitable vs vulnerable](/guides/exploitable-vs-vulnerable/) ## Sources - Fortinet advisory FG-IR-26-175 (FortiMail): https://fortiguard.fortinet.com/psirt/FG-IR-26-175 (read 19 Nov 2026) - NVD record for CVE-2026-104286: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-104286 (read 19 Nov 2026) - Palo Alto Networks advisory for CVE-2026-0322: https://security.paloaltonetworks.com/CVE-2026-0322 (read 19 Nov 2026) - NVD record for CVE-2026-0322: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-0322 (read 19 Nov 2026) - Cisco advisory cisco-sa-asaftd-tunnel-dos-s6mgkb75: https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-asaftd-tunnel-dos-s6mgkb75 (read 19 Nov 2026) - NVD record for CVE-2026-76418: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-76418 (read 19 Nov 2026) - BeyondTrust BT26-05 and ARMORY CASTLE: https://www.beyondtrust.com/blog/entry/bt26-05 (read 19 Nov 2026) - CISA KEV feed, dated 18 Nov 2026: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - CISA BOD 26-04: Prioritizing Security Updates Based on Risk: https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/triaging-an-internet-facing-appliance-finding/ --- # Agent investigation is only as checkable as the evidence record it leaves behind Guide, 19 Nov 2026, read 19 Nov 2026 **The short answer.** A checkable agent record names the finding, the access the agent held, each source with a timestamp, each test with its result, the conclusion and every unknown. A complete sample for CVE-2026-82329 in Artifactory follows, with a reviewer's replay steps and four vendors' own statements about showing evidence. ## A record has seven parts, and each part fails in its own way A conclusion alone is a claim. A record someone else can replay holds the parts below. The agent or the platform writes parts one to six, and a named person writes the seventh. The four-question tool on the home page sends a finding to agent investigation when a control sits in front of the service, or when two or three of code called, service reached and control in front are not sure. Those are the findings whose answer sits in configuration, logs and files, which is what this record holds. | Part | Written by | Usual failure | | --- | --- | --- | | 1. Finding and asset | Platform, from the scanner | Asset identity is a name that was reused | | 2. Access held | Platform, from the grant | Lists the role requested, not the access in force | | 3. Sources read, with time | Agent | Gives no time, so a stale read looks current | | 4. Tests against advisory preconditions | Agent | Cites no advisory text, or tests a different condition | | 5. Conclusion and scope | Agent | Says 'not exploitable' with no asset and date | | 6. Unknowns, with the reason | Agent | Omitted, so a gap reads as a pass | | 7. Review and expiry | Human reviewer | Missing, so the closure never expires | ## A complete sample record for CVE-2026-82329 on one Artifactory instance The question here is not only 'is the version affected' but 'was this instance compromised', which is why CVE-2026-82329 suits agent investigation. Fastly reports disclosure on 28 Aug 2026, about 171,000 attempts on 1 Sep 2026 and a peak near 406,000 on 2 Sep 2026. The instance, hosts, dates and counts in the record are an illustration. The advisory facts are real. Parts one and two come first. | Field | Entry | | --- | --- | | Record | INV-4417, version 1, opened 19 Nov 2026 09:12 UTC | | Finding | CVE-2026-82329, JFrog Artifactory Access service join-key authentication bypass, CVSS 9.8, added to KEV on 2 Sep 2026, first mass exploitation 1 Sep 2026 (Fastly) | | Asset | artifactory-prod-1, self-hosted, owner Platform Engineering | | Access held | Read-only API token with scope for system version and user listing; read-only search on the reverse-proxy log index. It holds no shell, file-system or write access | | Access granted by | Platform Engineering lead, 3 Nov 2026, expires 3 Feb 2027 | ## The sample record's tests each name a precondition, a source and a time The advisory preconditions are a self-hosted release in an affected range, a default install with no additionalJoinKeys set, and network access to the join endpoint. On a default install the Access service trusts an empty join key. The attack, as Fastly describes it, is a request to POST /access/api/v1/registry/join carrying a forged token whose key ID is the SHA-256 of the empty string, e3b0c442...b855. Fastly calls an HTTP 201 tied to that key ID "high-confidence evidence of compromise". | Test | Source, time (UTC) | Observed | Result | | --- | --- | --- | --- | | T1 Release affected now | GET system version, 19 Nov 2026 09:14 | 7.146.38 | Fixed build. JFrog's table lists 7.146.0 to 7.146.36 as affected and NVD treats every release below 7.146.38 as affected. 7.146.37 is in neither JFrog's affected list nor its patched list, so only 7.146.38 counts as fixed | | T2 Release affected during exposure | Change ticket CHG-9120, read 09:15 | Upgrade from 7.146.30 to 7.146.38 closed 9 Sep 2026 | Affected from 28 Aug to 9 Sep 2026 (12 days); 7.146.30 is inside both JFrog's and NVD's range | | T3 Join endpoint reachable | Reverse-proxy routing config as exposed in the log index, 09:17 | Route to /access/ open to the corporate VPN range only | Not internet reachable; reachable from the VPN range | | T4 Join requests with an HTTP 201 | Log index search, join path, status 201, 09:21 | Retention starts 21 Sep 2026. 0 matching requests from 21 Sep to 19 Nov 2026 | No evidence of compromise in the retained window | | T5 Unexpected admin users | User listing, 09:24 | 4 admins, all on the owner's roster | No unexpected admin | | T6 Groovy user plugins | Needs file-system access | Not run | Unknown | | T7 Join key rotated and tokens revoked since 28 Aug 2026 | Change tickets, search 'join key', 09:27 | No ticket found | Not done on record | ## The sample record's conclusion states its scope and lists its unknowns - Conclusion: artifactory-prod-1 is not vulnerable on its current release (T1). It was on an affected release from 28 Aug to 9 Sep 2026 (T2) and was reachable only from the VPN range (T3), so exposure to internet scanning was limited. - Compromise: unknown for 28 Aug to 20 Sep 2026, because the log index starts on 21 Sep 2026 (T4). No sign of compromise from 21 Sep 2026 onward (T4, T5). - Unknown, plugins: the agent holds no file-system access, so T6 was not run. - Open action for the reviewer: the ledger mitigation says patching does not revoke tokens already issued and advises rotating the platform join key and revoking tokens issued since 28 Aug 2026. T7 found no record that either was done. - Status: needs human action, not closed. Reviewer: Platform Engineering lead. Expiry of any later closure: 19 Feb 2027, or on a new advisory revision. - Version history: v1 only. A later version adds a row below with its date and the evidence that changed. ## A reviewer replays a record in five steps The same five steps work on any vendor's record. Run them on a record for a finding you already understand, and include one you expect to come back unknown. - Open the advisory and confirm the record quotes a precondition from it, not a paraphrase of the title. - For each test, repeat the source read yourself, or confirm the source and time are specific enough that you could. - Check the access list against the sources: every source read must fall inside the access held. T6 above shows a source outside it, correctly marked unknown. - Look for the unknowns. A record with no unknown on a finding whose evidence needs logs, files or hosts the agent cannot read has hidden a gap. - Confirm a person is named for the closure and that an expiry or trigger is set. The agent reads, and people decide on closures that become exceptions. ## Four vendors describe how their products show evidence These are the vendors' own statements, in alphabetical order. They describe design, not measured accuracy, and none comes with a sample record. - Cogent: "Every output includes the full chain of logic: data sources queried, agents involved, and why this path was selected over alternatives." When signal is short it says it "will surface unknowns/low confidence". - Maze: "Every recommendation comes with the evidence and reasoning behind it." - Onit: a Reachability Agent marks each exposure Reachable, Not Reachable or Unknown, and "when evidence is missing, an agent returns 'unknown' with the reason". - Seemplicity: AI Analysts investigate each asset by analyzing "live runtime configurations, exploit prerequisites, network reachability, code and dependency usage, and compensating controls". ## Sampling closures measures your own deployment Pull ten findings closed as not exploitable each month, read each record with the five steps above and count how many the reviewer reopens. That count is specific to your deployment. Keep the agent's access list under change control too, since a record that cites a source the agent can no longer reach cannot be replayed. ## Related pages - [Exploitable vs vulnerable](/guides/exploitable-vs-vulnerable/) - [Triaging an internet-facing appliance finding](/guides/triaging-an-internet-facing-appliance-finding/) - [How to prove a finding is not exploitable](/guides/prove-a-finding-is-not-exploitable/) ## Sources - Fastly on CVE-2026-82329, 3 Sep 2026: https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity (read 19 Nov 2026) - Cogent AI engine page: https://www.cogent.com/ai/cogent-ai-engine (read 19 Nov 2026) - Maze cloud platform page: https://mazehq.com/platform/cloud (read 19 Nov 2026) - Onit blog on AI agents in vulnerability management: https://onit.security/blog/ai-agents-vulnerability-management/ (read 19 Nov 2026) - Seemplicity AI Analysts announcement: https://seemplicity.ai/news/ai-analysts-autonomous-vulnerability-response/ (read 19 Nov 2026) - JFrog security advisories (CVE-2026-82329 affected and patched versions): https://docs.jfrog.com/releases/docs/jfrog-security-advisories (read 19 Nov 2026) - NVD record for CVE-2026-82329: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-82329 (read 19 Nov 2026) - CISA KEV entry for CVE-2026-82329 (added 2 Sep 2026): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-82329 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/guides/how-agent-investigation-records-its-evidence/ --- # 24 terms a reader meets, each with a sourced example Glossary, read 19 Nov 2026 **The short answer.** The glossary defines 24 terms used in exploitability work, from attack path and call graph to precondition and reachability. Standard terms quote NIST, FIRST, CISA, MITRE or the CVE Program; vendor descriptions are labelled as vendor claims and the site's own definitions as such. Each term adds a dated example from the 37 ledger CVEs. - [Agent investigation](#agent-investigation) - [Attack path](#attack-path) - [Attack vector](#attack-vector) - [Call graph](#call-graph) - [CVE](#cve) - [CVSS base score](#cvss-base) - [CWE](#cwe) - [Disclosure date](#disclosure-date) - [EPSS](#epss) - [Exploitability](#exploitability) - [Exploitation in the wild](#exploited-in-the-wild) - [Forensic triage](#forensic-triage) - [Indicator of compromise](#indicator-of-compromise) - [Known Exploited Vulnerabilities catalog](#kev) - [National Vulnerability Database](#nvd) - [Precondition](#precondition) - [Privileges required](#privileges-required) - [Proof of concept](#proof-of-concept) - [Reachability](#reachability) - [Runtime reachability](#runtime-reachability) - [Static reachability](#static-reachability) - [Transitive dependency](#transitive-dependency) - [Vulnerable](#vulnerable) - [Zero-day](#zero-day) ## Agent investigation No standards body defines agent investigation. Onit's documentation, a vendor claim, describes a Reachability Agent that "reads the network path an attacker would take, from the internet edge inward, and marks each exposure Reachable, Not Reachable or Unknown" ([Onit](https://onit.security/blog/ai-agents-vulnerability-management/)). On this site the term is the site's own: an automated reader that works one finding at a time and leaves a record of the evidence it used and what it could not settle. Example: CVE-2026-82329 in JFrog Artifactory is exploitable only on a default install where no additionalJoinKeys value is set and the Access service join endpoint is reachable, so an agent checks that setting, the exposure and tokens issued since the 28 Aug 2026 disclosure. See [the ledger entry](/ledger/cve-2026-82329/). ([source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories), read 19 Nov 2026) ## Attack path Microsoft's Defender for Cloud documentation, a vendor description of its own feature, says it "uses a proprietary algorithm to locate potential attack paths in your multicloud environment", and that these are paths that "start outside your organization and lead to critical targets" ([Microsoft Learn](https://learn.microsoft.com/en-us/azure/defender-for-cloud/how-to-manage-attack-path)). The site's own definition: an attack path joins reachability, privilege and controls into one route from an attacker's position to an asset. Example: CVE-2026-76460 in Cisco ISE, rated 10.0 and listed in KEV on 16 Sep 2026, turns on whether an attacker can reach the ISE API from a foothold, which is why Cisco's mitigation is access control lists. See [the ledger entry](/ledger/cve-2026-76460/). ([source](https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5), read 19 Nov 2026) ## Attack vector FIRST defines the CVSS Attack Vector metric as reflecting "the context by which vulnerability exploitation is possible", with the values network, adjacent, local and physical. The more remote the attacker can be, the higher the rating. Of the 37 ledger CVEs, 31 carry a network vector and 6 a local one. Example: CVE-2026-81963 in the Windows Update Stack carries AV:L and PR:L, so exploitation needs a low-privileged local session that already exists, and the 8 Sep 2026 cumulative updates fix it. See [the ledger entry](/ledger/cve-2026-81963/). (sources for the example: [NVD record for CVE-2026-81963](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-81963), [Microsoft CVRF for September 2026](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep), read 19 Nov 2026; [source for the definition](https://www.first.org/cvss/v4.0/specification-document), read 19 Nov 2026) ## Call graph Endor Labs, a vendor, describes its call graph analysis as a way to mitigate open source vulnerabilities by "pinpointing and understanding the invocation of vulnerable methods". A call graph maps which function calls which, so a tool can test whether any path from application code reaches the vulnerable function inside a dependency. Example: CVE-2026-16723 in fastjson 1.2.68 to 1.2.83 matters in a service only when a call path parses attacker-supplied JSON; the fix is fastjson 1.2.84, released 29 Jul 2026. See [the ledger entry](/ledger/cve-2026-16723/). ([source for the example](https://github.com/alibaba/fastjson/releases/tag/1.2.84), read 19 Nov 2026; [source for the definition](https://docs.endorlabs.com/scan/sca/call-graphs/index), read 19 Nov 2026) ## CVE The mission of the CVE Program, in its own words, is "to identify, define, and catalog publicly disclosed cybersecurity vulnerabilities". A CVE identifier names one flaw so that advisories, NVD and the KEV catalog can refer to the same thing. The identifier says nothing about severity, exposure or whether the flaw is exploited. Example: CVE-2026-111497 names the Brotli decoder out-of-bounds write that affects Node.js 22.x, 24.x and 26.x before the fixed releases, and any software embedding Brotli from 1.0.3 up to but excluding 1.0.10, 1.1.0 up to 1.1.1 and 1.2.0 up to 1.2.1. See [the ledger entry](/ledger/cve-2026-111497/). (sources for the example: [Node.js security releases of 21 Oct 2026](https://nodejs.org/en/blog/vulnerability/october-2026-security-releases), [Brotli 1.2.1 release notes](https://github.com/google/brotli/releases/tag/v1.2.1), [NVD record for CVE-2026-111497](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-111497), read 19 Nov 2026; [source for the definition](https://cve.mitre.org/), read 19 Nov 2026) ## CVSS base score FIRST states that the CVSS Base group "represents the intrinsic qualities of a vulnerability that are constant over time and across user environments". A base score therefore rates the flaw, not your system: it ignores whether the feature is on, the port is open or exploitation has been seen. Example: Broadcom rates CVE-2026-41709 in VMware ESX at 2.7, with high privileges required, and CISA still added it to KEV on 28 Oct 2026, 91 days after the 29 Jul 2026 disclosure. See [the ledger entry](/ledger/cve-2026-41709/). (sources for the example: [Broadcom advisory VMSA-2026-0006](https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017), [CISA KEV feed 2026.11.18](https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json), read 19 Nov 2026; [source for the definition](https://www.first.org/cvss/v4.0/specification-document), read 19 Nov 2026) ## CWE MITRE describes CWE as "a community-developed list of common software and hardware weaknesses", where a weakness is "a condition in a software, firmware, hardware, or service component that, under certain circumstances, could contribute to the introduction of vulnerabilities". A CWE names the type of flaw, not the instance. Example: CVE-2026-109618 in PHP is CWE-416, use after free, in PHP 8.2 before 8.2.35, 8.3 before 8.3.36, 8.4 before 8.4.27 and 8.5 before 8.5.12. See [the ledger entry](/ledger/cve-2026-109618/). (sources for the example: [PHP advisory GHSA-w8hw-w9fx-j79w](https://github.com/php/php-src/security/advisories/GHSA-w8hw-w9fx-j79w), [NVD record for CVE-2026-109618](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-109618), read 19 Nov 2026; [source for the definition](https://cwe.mitre.org/about/index.html), read 19 Nov 2026) ## Disclosure date Disclosure date is the site's own term: the day a vendor advisory or a CVE record first made the flaw public, whichever the ledger can date. Days from disclosure to first exploitation are counted from it, and a negative count means exploitation came first. Example: BeyondTrust disclosed CVE-2026-102764 on 15 Oct 2026 and reports exploitation from 29 Aug 2026, a gap of minus 47 days. Ten of the 33 were exploited on the disclosure day and two before it. See [the ledger entry](/ledger/cve-2026-102764/). ([source](https://www.beyondtrust.com/blog/entry/bt26-05), read 19 Nov 2026) ## EPSS FIRST states that "EPSS (Exploit Prediction Scoring System) is a data-driven model that estimates the probability a vulnerability will be exploited in the wild within the next 30 days", as a value between 0 and 1 for every CVE. It is a population-level estimate and not a measure of damage or of one environment. Example: FIRST's FAQ adds that "EPSS scores reflect exploitation probability across the global observed population, not your specific environment". On 18 Nov 2026 the EPSS API gave CVE-2026-76460 in Cisco ISE a probability of 0.756 (percentile 0.995) and CVE-2026-41709 in VMware ESX 0.0069 (percentile 0.535); both are in KEV, and ESX carries a CVSS base score of 2.7 ([FIRST EPSS API](https://api.first.org/data/v1/epss?cve=CVE-2026-41709,CVE-2026-76460), read 19 Nov 2026). None of the 37 ledger CVEs settles on an EPSS value alone. See [the ledger entry](/ledger/cve-2026-41709/). ([source](https://www.first.org/epss/faq), read 19 Nov 2026) ## Exploitability Exploitability, as the site defines it (no standard text applies): whether a flaw can be used against one specific system on one date, given its version, configuration, exposure and the attacker's starting access. The judgment belongs to an environment and is not a property of the CVE. A ledger entry lists every condition from the advisory that must hold at once. Example: CVE-2026-0322 in PAN-OS needs a build below the fixed build for its train, network reach to the management interface and a high-privilege account; the advisory is dated 14 Oct 2026. See [the ledger entry](/ledger/cve-2026-0322/). ([source](https://security.paloaltonetworks.com/CVE-2026-0322), read 19 Nov 2026) ## Exploitation in the wild CISA's KEV page defines a vulnerability under active exploitation as "one for which there is reliable evidence that execution of malicious code was performed by an actor on a system without permission of the system owner", and says this includes attempted as well as successful exploitation. A vendor advisory may report exploitation earlier or later than the catalog does. Example: Fortinet advisory FG-IR-26-175, published 1 Oct 2026, says CVE-2026-104286 "has been reported to be exploited in the wild", and CISA listed it the same day. See [the ledger entry](/ledger/cve-2026-104286/). ([source for the example](https://fortiguard.fortinet.com/psirt/FG-IR-26-175), read 19 Nov 2026; [source for the definition](https://www.cisa.gov/known-exploited-vulnerabilities), read 19 Nov 2026) ## Forensic triage CISA's BOD 26-04 says forensic triage means that the agency must "complete remediation or mitigation action within the timeline (three days)" and also triage the asset "to assess whether the system is compromised". Patching alone does not show whether an attacker was already inside. Example: In KEV feed 2026.11.18, 19 of the 33 ledger CVEs carry the forensic triage flag, among them CVE-2026-0322 and CVE-2026-76460, while CVE-2026-41709 does not. See [the ledger entry](/ledger/cve-2026-76460/). (sources for the example: [CISA KEV feed 2026.11.18](https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json), read 19 Nov 2026; [source for the definition](https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk), read 19 Nov 2026) ## Indicator of compromise NIST's glossary defines indicators of compromise as "Technical artifacts or observables that suggest that an attack is imminent or is currently underway or that a compromise may have already occurred". Checking for them after patching shows whether someone got in before the fix. Example: Fortinet lists two indicator addresses, 79[.]141[.]169[.]187 and 45[.]129[.]0[.]192, and system event log lines for CVE-2026-104286 in FG-IR-26-175. See [the ledger entry](/ledger/cve-2026-104286/). ([source for the example](https://fortiguard.fortinet.com/psirt/FG-IR-26-175), read 19 Nov 2026; [source for the definition](https://csrc.nist.gov/glossary/term/indicator_of_compromise), read 19 Nov 2026) ## Known Exploited Vulnerabilities catalog CISA's KEV catalog page says CISA "maintains the authoritative source of vulnerabilities that have been exploited in the wild". An entry shows that exploitation was observed somewhere, not that every installation is exposed. Each entry carries a due date, a ransomware flag and, under BOD 26-04, a forensic triage flag. Example: The JSON feed of version 2026.11.18, read on 19 Nov 2026, held 1,786 entries, and 33 of the 37 ledger CVEs are in it, matched by CVE ID. CVE-2026-76418 in Cisco ASA and FTD was added on 29 Oct 2026. See [the ledger entry](/ledger/cve-2026-76418/). (sources for the example: [CISA KEV feed 2026.11.18](https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json), read 19 Nov 2026; [source for the definition](https://www.cisa.gov/known-exploited-vulnerabilities-catalog), read 19 Nov 2026) ## National Vulnerability Database The National Vulnerability Database, run by NIST, publishes each CVE record with a CVSS rating, a weakness type and affected-product ranges expressed as CPE data. Its version ranges can differ from the vendor's table, so a ledger record gives the vendor's table where the two disagree and says so. Example: The NVD record for CVE-2026-46300 lists Linux 3.9 through 5.10.257 as affected, but no 5.10 stable release from 5.10.258 to 5.10.274 contains the fix, which the ledger record states. See [the ledger entry](/ledger/cve-2026-46300/). ([source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-46300), read 19 Nov 2026) ## Precondition Precondition is the site's own term: a condition in an advisory that must hold before a flaw can be exploited, such as a version, an enabled feature, a reachable interface or an attacker privilege. A finding is exploitable on a system only when all of its preconditions hold. Ledger entries list preconditions before the fix. Example: CVE-2026-123158 needs sshd built against musl libc without PAM, password authentication on, a reachable port, and an account with a * password field whose name the attacker knows. OpenSSH 10.6p2 shipped on 5 Nov 2026. See [the ledger entry](/ledger/cve-2026-123158/). ([source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-123158), read 19 Nov 2026) ## Privileges required FIRST says this metric "describes the level of privileges an attacker must possess prior to successfully exploiting the vulnerability", with the values none, low and high. Across the 37 ledger CVEs, 25 need none, 10 need low privileges and 2 need high privileges. Example: CVE-2026-0322 in PAN-OS and CVE-2026-41709 in VMware ESX are the two entries with PR:H, which means an administrator account on the management interface or host. See [the ledger entry](/ledger/cve-2026-0322/). (sources for the example: [NVD record for CVE-2026-0322](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-0322), [NVD record for CVE-2026-41709](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-41709), read 19 Nov 2026; [source for the definition](https://www.first.org/cvss/v4.0/specification-document), read 19 Nov 2026) ## Proof of concept CISA's KEV page defines proof of concept (PoC) as "the code for a vulnerability that, when executed, would allow for exploitation", and says that "making PoC publicly available can increase the likelihood of an attacker exploiting the vulnerability in the wild" but does not by itself show that the flaw has been or will be exploited. The ledger records that a proof of concept appeared, with its date, and never reproduces its steps. Example: Fastly counted about 75,000 attempts against Artifactory CVE-2026-82329 on 31 Aug 2026, mostly offensive security services, and about 171,000 on 1 Sep 2026 after a public exploit appeared. See [the ledger entry](/ledger/cve-2026-82329/). ([source for the example](https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity), read 19 Nov 2026; [source for the definition](https://www.cisa.gov/known-exploited-vulnerabilities), read 19 Nov 2026) ## Reachability The OpenVEX specification names the case that reachability analysis tests, the status justification vulnerable_code_not_in_execute_path: "Typically, this case occurs when the product includes the vulnerable subcomponent but does not call or use the vulnerable code". The question has two halves: is the code called, and can the service be reached from where an attacker stands. Example: Snyk, a vendor, says its reachability analysis works by "identifying whether your application is calling a code element (for example, functions, classes, modules, and more) related to the vulnerability", and notes that "it is not always possible to give a precise answer when reflection programming is used", so no path found is weaker than path ruled out. CVE-2026-111497 is reachable only where a service decompresses Brotli data an attacker can influence. See [the ledger entry](/ledger/cve-2026-111497/). ([source for the example](https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis), read 19 Nov 2026; [source for the definition](https://github.com/openvex/spec/blob/main/OPENVEX-SPEC.md), read 19 Nov 2026) ## Runtime reachability Endor Labs, a vendor, describes instrumented container reachability as embedding "a runtime sensor in your container image to record how your application uses the image in a real environment". Runtime reachability observes calls while a workload runs, so it proves a call happened and says nothing about paths the observation window never exercised. Example: CVE-2026-46300 in the Linux kernel is settled on the running host: whether esp4 or esp6 is loaded and whether unprivileged user namespaces are allowed. The stable kernel releases carrying the fix shipped on 23 May 2026. See [the ledger entry](/ledger/cve-2026-46300/). ([source for the example](https://access.redhat.com/security/cve/CVE-2026-46300), read 19 Nov 2026; [source for the definition](https://docs.endorlabs.com/scan/containers/instrumented-reachability/index), read 19 Nov 2026) ## Static reachability Static reachability is the site's own term; no standard text defines it. It reads code or configuration without running it, and asks whether a path runs from an entry point to the vulnerable element or whether the affected version and setting are present. A found path shows exposure is possible. A missing path is weaker evidence, because reflection and dynamic loading hide calls. Example: CVE-2026-94545 affects Satori and next/og only on the Node.js runtime and only when attacker-controlled values reach SVG content, attributes or styles; a source search decides it, and Next.js 16.3.6 fixes it. See [the ledger entry](/ledger/cve-2026-94545/). ([source](https://github.com/vercel/next.js/security/advisories/GHSA-vcvr-r3jv-pc5j), read 19 Nov 2026) ## Transitive dependency The Apache Maven documentation says Maven "avoids the need to discover and specify the libraries that your own dependencies require by including transitive dependencies automatically". A scanner can flag a flaw in a library the manifest never names, while the call site sits two levels down. The resolved version in the build output, not the manifest, shows which version is present. Example: CVE-2026-16723 affects fastjson 1.2.68 to 1.2.83 and is fixed in 1.2.84. A Maven project that never lists com.alibaba:fastjson in its pom.xml still carries it when another dependency requires it, and mvn dependency:tree shows the resolved version. See [the ledger entry](/ledger/cve-2026-16723/). ([source for the example](https://github.com/alibaba/fastjson/releases/tag/1.2.84), read 19 Nov 2026; [source for the definition](https://maven.apache.org/guides/introduction/introduction-to-dependency-mechanism.html), read 19 Nov 2026) ## Vulnerable NIST's glossary defines a vulnerability as a "Weakness in an information system, system security procedures, internal controls, or implementation that could be exploited or triggered by a threat source." A system is vulnerable when it runs affected code, which says nothing yet about whether an attacker can reach it. Example: CVE-2026-75604 affects Next.js 13.4.0 before 15.5.24 and 16.0.0 before 16.3.3, but the NVD record names a Windows filesystem as a condition. A Linux host running 15.5.20 falls inside the version ranges, so a scanner flags it, yet the record's Windows condition does not hold there. A version match alone does not settle it. See [the ledger entry](/ledger/cve-2026-75604/). ([source for the example](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604), read 19 Nov 2026; [source for the definition](https://csrc.nist.gov/glossary/term/vulnerability), read 19 Nov 2026) ## Zero-day NIST's glossary, quoting CNSSI 4009-2022, defines a zero-day attack as "An attack that exploits a previously unknown hardware, firmware, or software vulnerability." In the ledger, exploitation dated before the disclosure date marks a zero-day, and two entries meet that test. Example: CVE-2026-75650 in Adobe Commerce was exploited from 4 Sep 2026, three days before Adobe's bulletin of 7 Sep 2026. CVE-2026-102764 in BeyondTrust was exploited 47 days before disclosure. See [the ledger entry](/ledger/cve-2026-75650/). (sources for the example: [Sansec research](https://sansec.io/research/stylesmuggler-0day), [BeyondTrust bulletin BT26-05](https://www.beyondtrust.com/blog/entry/bt26-05), read 19 Nov 2026; [source for the definition](https://csrc.nist.gov/glossary/term/zero_day_attack), read 19 Nov 2026) ## Related pages - [Exploitability Ledger: the data, downloadable](/datasets/) - [The Methods Matrix](/matrix/) ## Sources - https://docs.jfrog.com/releases/docs/jfrog-security-advisories (read 19 Nov 2026) - https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5 (read 19 Nov 2026) - https://www.first.org/cvss/v4.0/specification-document (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-81963 (read 19 Nov 2026) - https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep (read 19 Nov 2026) - https://docs.endorlabs.com/scan/sca/call-graphs/index (read 19 Nov 2026) - https://github.com/alibaba/fastjson/releases/tag/1.2.84 (read 19 Nov 2026) - https://cve.mitre.org/ (read 19 Nov 2026) - https://nodejs.org/en/blog/vulnerability/october-2026-security-releases (read 19 Nov 2026) - https://github.com/google/brotli/releases/tag/v1.2.1 (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-111497 (read 19 Nov 2026) - https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017 (read 19 Nov 2026) - https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - https://cwe.mitre.org/about/index.html (read 19 Nov 2026) - https://github.com/php/php-src/security/advisories/GHSA-w8hw-w9fx-j79w (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-109618 (read 19 Nov 2026) - https://www.beyondtrust.com/blog/entry/bt26-05 (read 19 Nov 2026) - https://www.first.org/epss/faq (read 19 Nov 2026) - https://security.paloaltonetworks.com/CVE-2026-0322 (read 19 Nov 2026) - https://www.cisa.gov/known-exploited-vulnerabilities (read 19 Nov 2026) - https://fortiguard.fortinet.com/psirt/FG-IR-26-175 (read 19 Nov 2026) - https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk (read 19 Nov 2026) - https://csrc.nist.gov/glossary/term/indicator_of_compromise (read 19 Nov 2026) - https://www.cisa.gov/known-exploited-vulnerabilities-catalog (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-46300 (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-123158 (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-0322 (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-41709 (read 19 Nov 2026) - https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity (read 19 Nov 2026) - https://github.com/openvex/spec/blob/main/OPENVEX-SPEC.md (read 19 Nov 2026) - https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis (read 19 Nov 2026) - https://docs.endorlabs.com/scan/containers/instrumented-reachability/index (read 19 Nov 2026) - https://access.redhat.com/security/cve/CVE-2026-46300 (read 19 Nov 2026) - https://github.com/vercel/next.js/security/advisories/GHSA-vcvr-r3jv-pc5j (read 19 Nov 2026) - https://maven.apache.org/guides/introduction/introduction-to-dependency-mechanism.html (read 19 Nov 2026) - https://csrc.nist.gov/glossary/term/vulnerability (read 19 Nov 2026) - https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604 (read 19 Nov 2026) - https://csrc.nist.gov/glossary/term/zero_day_attack (read 19 Nov 2026) - https://sansec.io/research/stylesmuggler-0day (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/glossary/ --- # The Exploitability Ledger and the Methods Matrix, downloadable Dataset edition 19 Nov 2026 **The short answer.** Two datasets are published as CSV and JSON: the Exploitability Ledger with 37 headline CVEs disclosed from 9 Apr to 5 Nov 2026 (CVSS, KEV and exploitation dates, exposure conditions, settling method) and the [Methods Matrix](/matrix/) with seven methods by question, need, proof, blind spot and provider. Both carry the read date of 19 Nov 2026. Each dataset is a CSV file and a JSON file with the same rows. CSV files have one header row and comma-separated cells. JSON files hold the rows under a rows key, with the dataset name, edition, read date, source, license, cite line and description. All dates are ISO 8601, written YYYY-MM-DD. Fixes, mitigations and detection steps are on each ledger page and not in the files. The column terms (CVSS base score, KEV due date, attack vector) are defined in the [glossary](/glossary/). Each ledger row also has its own page, such as [CVE-2026-76460 in Cisco ISE](/ledger/cve-2026-76460/). Ledger fields, one row per CVE. cve: the CVE identifier. name: the short flaw name used on the ledger page. vendor and product: as named in the vendor advisory. cvss: the CVSS base score, written with one decimal, from the CNA that issued the CVE or, where it gave none, from NVD. cvss_version: 3.1 or 4.0, the prefix of the vector shown on the ledger page; 29 rows are 3.1 and 8 are 4.0. disclosed: the advisory date, or the NVD publication date when the advisory has none. kev_added: the date CISA added the CVE to the KEV catalog, feed version 2026.11.18, empty when the CVE is not listed. kev_due: the due date in the same feed, empty when not listed. first_exploited: the earliest dated report of exploitation from the vendor, a named research firm or CISA; when only a CISA date exists it is an upper bound, the latest date exploitation can have started. first_exploited_basis: where that date comes from, one of named report, CISA date (KEV) for the KEV addition, or CISA date (SSVC) for CISA's SSVC entry in the NVD record; empty when no report is dated. days_to_exploitation: whole calendar days from disclosed to first_exploited, where 0 is the same day, a negative number means exploitation was reported before disclosure, and an empty cell means no dated report exists. method_that_settles_it: the method of the matrix that settles whether the flaw is exploitable in a given environment. preconditions: the conditions from the ledger page's exposure table, all of which must hold, joined with ' | '. Sample ledger row (ledger.csv): `CVE-2026-76460,Cisco ISE API authentication bypass,Cisco,Identity Services Engine,10.0,3.1,2026-09-16,2026-09-16,2026-09-19,2026-09-16,CISA date (KEV),0,Attack-path analysis,"a node runs a release below its fixed patch, or 3.0 | the attacker can reach the node's API"` Matrix fields, one row per method, alphabetical by name. method: the method name. slug: the last part of the method page address. question: the question the method answers. needs, proves, misses: the three statements on the method page. offered_by: the providers of the method, named alphabetically and separated by commas. ledger_entries_settled: how many of the 37 ledger rows name this method in method_that_settles_it. The counts are: static reachability 14, attack-path analysis 11, runtime reachability 7, agent investigation 3, known-exploited lists 2, severity scores 0 and EPSS exploit prediction 0. Both datasets, in CSV and JSON, are licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/): reuse them with the cite line below. ## Downloads - [Exploitability Ledger (CSV)](/datasets/ledger.csv): 37 rows, edition 19 Nov 2026 - [Exploitability Ledger (JSON)](/datasets/ledger.json): 37 rows, edition 19 Nov 2026, with the cite line - [Methods Matrix (CSV)](/datasets/matrix.csv): 7 rows, edition 19 Nov 2026 - [Methods Matrix (JSON)](/datasets/matrix.json): 7 rows, edition 19 Nov 2026, with the cite line ## Cite this **Cite this.** Reachability Analysis, Exploitability Ledger and Methods Matrix, edition 19 Nov 2026 (first edition 10 Nov 2026), https://reachabilityanalysis.com/datasets/ ## Changelog - 10 Nov 2026: first edition, published as Is It Exploitable with 40 ledger CVEs. - 19 Nov 2026: second edition. All 37 ledger records were re-read against NVD, the CISA KEV feed 2026.11.18 and the vendor advisory; three records were removed; the ledger range is 9 Apr to 5 Nov 2026; the cite line points to this page. New columns: cvss_version, kev_due and first_exploited_basis in the ledger, question and ledger_entries_settled in the matrix. cvss is written with one decimal, and preconditions lists the exposure conditions only. ## Related pages - [The Methods Matrix](/matrix/) - [Glossary](/glossary/) ## Sources - NIST National Vulnerability Database CVE API 2.0 (per-CVE records, cveId parameter): https://services.nvd.nist.gov/rest/json/cves/2.0 (read 19 Nov 2026) - CISA Known Exploited Vulnerabilities catalog feed, version 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - Cisco advisory cisco-sa-ISE-ABP-VNSW7Tn5 (CVE-2026-76460), the source of the sample row: https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5 (read 19 Nov 2026) - Creative Commons Attribution 4.0 International (the license of both datasets): https://creativecommons.org/licenses/by/4.0/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/datasets/ --- # About Reachability Analysis About, updated 19 Nov 2026 **The short answer.** Reachability Analysis is a reference on one question: can a given finding be exploited where you run it. The second edition, dated 19 Nov 2026, holds a ledger of 37 headline CVEs disclosed from 9 Apr to 5 Nov 2026, seven methods compared and guides, each with a source and a read date. The reader in mind is the analyst or engineer holding a scanner finding and a deadline. A severity score says how bad a flaw could be. The question here is narrower: do the conditions that the advisory names exist where you run the software? Two original assets answer it. The Exploitability Ledger gives, for each of 37 headline CVEs, the preconditions, the affected and fixed versions, the mitigation, the detection and the method that settles exposure: 14 settle on static reachability, 11 on attack-path analysis, 7 on runtime reachability, 3 on agent investigation and 2 on known-exploited evidence. Of the 37, 33 are in the CISA KEV catalog with a dated exploitation report. The Methods Matrix sets seven methods side by side on what each needs, what it proves, what it misses and who offers it. The guides, listed on the home page, show how to apply both to a finding in your own backlog, using ledger CVEs as worked examples. Facts come from vendor advisories, the National Vulnerability Database and the CISA Known Exploited Vulnerabilities catalog. How they are sourced, labelled and ordered is set out once, on the [editorial policy](/editorial-policy/) page. The first edition appeared on 10 Nov 2026 under the title Is It Exploitable. The second edition, on 19 Nov 2026, uses the title Reachability Analysis. It re-read all 37 records, removed three, and folded 24 ledger entry pages and several guide, method and news pages into the home page, the Methods Matrix or the editorial policy, each answering with a 301. The rows of the folded entries stay in the datasets. Reach the editor at the address on the [contact](/contact/) page. A reader who finds an error is asked to name the page, the claim and a source. ## Related pages - [Editorial policy](/editorial-policy/) - [Contact](/contact/) - [Privacy](/privacy/) - [Terms](/terms/) ## Sources - CISA Known Exploited Vulnerabilities catalog feed, version 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/about/ --- # How Reachability Analysis is written Editorial policy, updated 19 Nov 2026 **The short answer.** Editorial policy: each fact on Reachability Analysis carries a source address and the date the editor read that source, 19 Nov 2026 for every record in the second edition. CVE facts come from the vendor advisory, NVD and the CISA Known Exploited Vulnerabilities catalog. The site lists and does not rank. A ledger record takes its preconditions, affected and fixed versions, mitigation and detection from the vendor advisory. The disclosure date is the advisory date, or the NVD publication date when the advisory has none. The KEV date comes from the CISA feed. The first-exploited date is the earliest dated report of exploitation, from one of four bases, listed here from the most direct to the least: a vendor advisory or release note, a named research firm, the date CISA's SSVC entry in the NVD record marks exploitation as active, and the date CISA added the CVE to KEV. The ledger labels the first two a named report and the last two a CISA date. A CISA date is an upper bound: exploitation cannot have started later than that date, and may have started earlier. When a record has only the KEV addition, its first-exploited date equals the KEV date; the SSVC date can be earlier than the KEV date, as for CVE-2026-46300, where it is 14 May 2026 and the KEV date is 28 Oct 2026. A statement about what a vendor's product does is labelled the vendor's claim and cites the vendor page. A statement from an outside party is labelled a third-party source. Statements about a method come from its publisher: FIRST for CVSS and [EPSS](/methods/exploit-prediction/), CISA for the KEV catalog, and vendor documentation for reachability and attack paths. The editor opens each source before a fact is published, and a claim whose source cannot be opened is left out. A read date is the day the editor opened the source and is shown beside the link. It is not the date the advisory was published or revised, which the record gives separately when the source states it. An empty exploitation field means no dated report of exploitation existed on the read date. Every record in the second edition was re-read on 19 Nov 2026 against NVD, the CISA KEV feed (version 2026.11.18) and the vendor advisory. When a source changes, the record is revised in place. Order follows a fixed rule on each page: vendors are alphabetical, CVE records run newest first by disclosure date, and methods are alphabetical. Order never reflects merit, danger or importance. No exploit code, proof-of-concept steps, payloads or scanning instructions appear anywhere on the site. No page grades a vendor or a method, puts them in an order of merit or names a favorite. CVSS base scores are the ones published in the CVE record by the CVE Numbering Authority that issued the CVE or, where it gave none, by NVD; EPSS values are the ones FIRST publishes. A vendor may send a newer source for a claim about its own listed method. The record changes to the new page with a new read date, and the change is logged. Guides are written for a working analyst, use ledger CVEs as worked examples and end with their sources. ## Related pages - [About](/about/) - [Contact](/contact/) - [Privacy](/privacy/) - [Terms](/terms/) ## Sources - CISA Known Exploited Vulnerabilities catalog: https://www.cisa.gov/known-exploited-vulnerabilities-catalog (read 19 Nov 2026) - FIRST CVSS v4.0 specification: https://www.first.org/cvss/v4.0/specification-document (read 19 Nov 2026) - FIRST EPSS frequently asked questions: https://www.first.org/epss/faq (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/editorial-policy/ --- # Contact Reachability Analysis Contact, updated 19 Nov 2026 **The short answer.** Contact the editor at [editor@reachabilityanalysis.com](mailto:editor@reachabilityanalysis.com) by email. Include the page, the claim and a primary source such as a vendor advisory, an NVD entry or a CISA Known Exploited Vulnerabilities listing. A confirmed correction is made on the page itself. Three kinds of message are useful. The first is a correction: quote the line as printed, say what is wrong and give one address that shows the right fact. The second is a news tip, with a link to the primary page, such as an advisory or a catalog entry. The third is a request for the datasets. The Exploitability Ledger and the Methods Matrix are published as CSV and JSON on the [home page](/). If you want a different cut of them, say which fields you need. Reuse is under CC BY 4.0 with the cite line printed on that page. To ask for a CVE to be added to the ledger, send the CVE identifier and the vendor advisory. The ledger holds 37 headline CVEs disclosed from 9 Apr to 5 Nov 2026, so say why the bug belongs among them. A request is not a promise to add it. Questions about how a record was built are welcome. The answer points to the advisory, the NVD entry and the KEV entry the record draws on, with the read date. Vendors may send a factual correction to their own listed claims, with a link to the page that supports it. Please do not send exploit code, proof-of-concept material or scan results. If you have found a flaw in a product, report it to the vendor named on that product's advisory. The site has no forms, accounts or mailing list, so email is the only way to reach the editor. ## Related pages - [About](/about/) - [Editorial policy](/editorial-policy/) - [Privacy](/privacy/) - [Terms](/terms/) ## Sources - Creative Commons Attribution 4.0 International: https://creativecommons.org/licenses/by/4.0/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/contact/ --- # Privacy Privacy, updated 19 Nov 2026 **The short answer.** Privacy: Reachability Analysis sets no cookies, runs no analytics and loads no third-party scripts. It has no forms, accounts or mailing list, and its fonts are self-hosted. The only record of a visit is the access log of the delivery network that serves the pages, kept for 400 days. Edition of 19 Nov 2026. The pages are static HTML files on Amazon S3, delivered by Amazon CloudFront. Tools on the site run in your browser, and the values you enter are not sent anywhere. CloudFront standard access logs go to a private S3 bucket in the US East (N. Virginia) region. A log line holds the date and time, the edge location, the client IP address, the requested path, the status code, the referrer and the user agent. The [AWS standard logging reference](https://docs.aws.amazon.com/AmazonCloudFront/latest/DeveloperGuide/standard-logs-reference.html) lists these fields (read 19 Nov 2026). The logs are used for two counts, page views per page and visits by crawlers. A lifecycle rule on the bucket, set on 19 Nov 2026, deletes log files after 400 days. The site does not build a profile of any reader and does not use the log to show different content to different people. The controller of the log is Reachability Analysis, reachable at the address below. The basis for keeping it is the legitimate interest in operating and securing the site. The client IP address is personal data in many jurisdictions, and it sits in the log for up to 400 days. The site matches log lines to no other record of a reader, because it keeps none. Email sent to editor@reachabilityanalysis.com is kept to answer it and to log any correction that follows. To ask about the log lines tied to an address, or to have an email message deleted, write to the same address. No data is sold. The site shows no advertising and loads nothing from an advertising network, so no reader data passes to one through the pages. ## Related pages - [About](/about/) - [Editorial policy](/editorial-policy/) - [Contact](/contact/) - [Terms](/terms/) ## Sources - Amazon CloudFront standard log file fields: https://docs.aws.amazon.com/AmazonCloudFront/latest/DeveloperGuide/standard-logs-reference.html (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/privacy/ --- # Terms of use Terms, updated 19 Nov 2026 **The short answer.** Terms of use: Reachability Analysis is published for general information, and the current edition is dated 19 Nov 2026. Exploitability depends on your own systems, so confirm each record against the vendor advisory before acting. Datasets may be reused under CC BY 4.0 with the cite line below. The cite line is: Reachability Analysis, Exploitability Ledger and Methods Matrix, edition 19 Nov 2026, https://reachabilityanalysis.com/datasets/ledger.csv. Name the edition you used, because advisories get revised and records change with them. Every record links its primary source so you can read the current state. The site gives general information and not legal advice or an incident response service. A severity score, a catalog listing or a ledger row is one input. The decision to patch, mitigate or accept a risk remains with the people who run the system, and the editor accepts no liability for decisions made on the strength of the site. Do not present the datasets as an assessment of your environment. They describe public advisories and public catalog entries. They cannot say whether a particular version, configuration or network path exists in your estate. Product and company names belong to their owners and appear so a reader can tell which product a record is about. Quotations from advisories, standards and vendor pages stay the property of their authors. Short passages of the site's text, up to 100 words, may be quoted with a link back to the page. Longer reuse of the text, including the guides, needs permission by email. The datasets are the exception: CC BY 4.0 allows copying and adapting them, including for commercial use, if the cite line is kept. Links to other sites are given so you can read the source. The text on those sites is theirs and can change at any time. Before you rely on a record for a purchase, an exception or a compliance decision, open its source. If the source and the record differ, write to editor@reachabilityanalysis.com and the record will be corrected and logged. ## Related pages - [About](/about/) - [Editorial policy](/editorial-policy/) - [Contact](/contact/) - [Privacy](/privacy/) ## Sources - Creative Commons Attribution 4.0 International: https://creativecommons.org/licenses/by/4.0/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/terms/ --- # CVE-2026-123158: OpenBSD OpenSSH, is it exploitable where you run it Ledger entry, disclosed 5 Nov 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-123158 affects OpenBSD OpenSSH and was disclosed 5 Nov 2026 with a CVSS score of 8.1. Exposure needs these conditions: sshd is OpenSSH before 10.6p2 on Linux with musl libc and no PAM; PasswordAuthentication is yes, the SSH port is reachable, and one account has a * password field while another has a $5$ or $6$ hash. ## What the flaw is When sshd is built without PAM it checks a password by hashing the attempt with crypt() and comparing the result with the stored password field. For an account whose field is too short to hold a salt, such as the * that marks an account with no password, sshd borrows the salt of another account. That code dates from the 2016 fix for CVE-2016-6210. musl's crypt() returns the one-character string * when it fails, and it fails for any password longer than 256 bytes. The stored field of a no-password account is also *. The two strings match, so the login succeeds. The researcher reports that the bypass also depends on sshd reading its password field from a buffer that libc reuses, so it needs more than one attempt on a connection. An attacker who can reach the SSH port and knows an account name with a * field logs in without a password. One account with a real SHA-256 or SHA-512 hash makes every no-password account reachable. On Alpine cloud images the alpine user has passwordless doas, so the login becomes root. This page gives no exploit strings. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | sshd is OpenSSH before 10.6p2 on Linux with musl libc and no PAM | Run `ssh -V`. On Alpine run `apk info -v openssh-server` and compare it with the fixed package table. A distribution package can carry the fix under an older version string. musl: `ldd $(command -v sshd)` shows libc.musl (on Alpine, `cat /etc/os-release` names the distribution). No PAM: `ldd $(command -v sshd) \| grep -i pam` prints nothing, or on Alpine `apk info -e openssh-server-pam` finds no package, or `sshd -T \| grep -i usepam` shows no. | | PasswordAuthentication is yes, the SSH port is reachable, and one account has a * password field while another has a $5$ or $6$ hash | Password authentication: `sshd -T \| grep -i passwordauthentication` shows yes, the default; check Match blocks in /etc/ssh/sshd_config and /etc/ssh/sshd_config.d/ too. A * account: `awk -F: '$2=="*"{print $1}' /etc/shadow`; also count directory-service users through musl-nscd (nslcd, winbind, sssd), which have no shadow entry, and note that accounts locked with ! are not affected. A hashed account: `awk -F: '$2 ~ /^\$[56]\$/{print $1}' /etc/shadow` prints names only, and an empty result ends the exposure. Reachable port: `ss -ltnp \| grep sshd` and the firewall and security-group rules for port 22; names such as root or alpine are easy to guess. | | Root or an escalation path is available through the * account. | Run `sshd -T \| grep -i permitrootlogin`, and check /etc/doas.conf and /etc/doas.d/ for passwordless rules. Alpine tiny-cloud images from 3.18.11 through 3.24.3 ship the alpine user with passwordless doas. | ## Affected and fixed versions | Product or line | Affected | First fixed | | --- | --- | --- | | OpenSSH portable (Linux, musl, no PAM) | Before 10.6p2 (NVD: before 10.6, 10.6, 10.6p1) | 10.6p2 (5 Nov 2026) | | Alpine edge | openssh before 10.6_p2-r0 | 10.6_p2-r0 | | Alpine 3.24 | openssh before 10.3_p1-r2 | 10.3_p1-r2 (point release 3.24.4) | | Alpine 3.23 | openssh before 10.2_p1-r1 | 10.2_p1-r1 (point release 3.23.8) | | Alpine 3.22 | openssh before 10.0_p1-r11 | 10.0_p1-r11 (point release 3.22.8) | | Alpine 3.21 (end of support 1 Nov 2026 per the Alpine releases page) | openssh before 9.9_p2-r1 | 9.9_p2-r1 (point release 3.21.10) | | Alpine 3.20 | openssh before 9.7_p1-r6 | 9.7_p1-r6 (package only) | | Alpine 3.19 | openssh before 9.6_p1-r3 | 9.6_p1-r3 (package only) | | Alpine 3.18 | openssh before 9.3_p2-r4 | 9.3_p2-r4 (package only) | | Alpine 3.17 | openssh before 9.1_p1-r7 | 9.1_p1-r7 (package only) | | Alpine 3.16 | openssh before 9.0_p1-r6 | 9.0_p1-r6 (package only) | | Alpine tiny-cloud images (qcow2, VHD, AMI) | 3.18.11 through 3.24.3 | Images from 3.21.10, 3.22.8, 3.23.8 and 3.24.4 and later; older images need an openssh update on each running instance | | Linux with glibc, or PAM in the auth path, or non-Linux | Not affected (OpenSSH) | Not applicable | Alpine's list covers every release back to 3.16. Other musl images (OpenWrt, Buildroot, Yocto) need the 10.6p2 source or the fix commit, which Medcrypt reports applies to OpenSSH 7.3p1 and later. ## If the fix cannot go in yet - Upgrade to OpenSSH 10.6p2 or the Alpine fixed package (apk upgrade -U -a), then restart sshd with rc-service sshd restart. Cost: a short sshd restart; existing sessions keep running - If you cannot upgrade, set PasswordAuthentication no in sshd_config, including inside Match blocks, check with sshd -t, restart sshd and confirm with sshd -T | grep -i passwordauthentication. OpenSSH and Alpine both give this as the immediate step. Cost: nobody can log in with a password. Test key login in a second session first and keep the first session open - OpenSSH's own wording for musl systems without PAM is to set PasswordAuthentication=no or disable SSH immediately. Disabling SSH cuts all remote administration of that host - Medcrypt, a third party, reports that MaxAuthTries 1 keeps password login working while blocking its reproduction, and that MaxAuthTries 2 does not. Treat it as a stopgap only. Cost: a user who mistypes a password once is disconnected - Images that cannot take 10.6p2: Medcrypt reports the fix commit applies cleanly to OpenSSH 7.3p1 and later. Cost: a rebuild of the image - CISA's KEV entry requires applying the vendor mitigations under BOD 26-04, with forensic triage, by 12 Nov 2026 ## What to look for - Alpine: search /var/log/messages for lines reading Accepted password for an account that has no password (alpine, root or any key-only user). Then review doas use, recently created users, new files and authorized_keys entries. If you cannot rule out a login, Alpine says to treat the machine as compromised - GreyNoise indicators from intrusions it observed from 7 Nov 2026: scanning and exploitation from 183[.]220[.]91[.]15, 20[.]255[.]63[.]133, 209[.]248[.]50[.]17, 79[.]116[.]138[.]30, 20[.]2[.]81[.]241 and 47[.]80[.]65[.]43; a command-and-control domain tac[.]vpnla[.]org pinned in /etc/hosts as 104[.]21[.]88[.]83; a Python runtime under /opt/py3 unpacked from /tmp/py3.tar.gz; an agent service script at /etc/init.d/tac-agent (SHA-256 7cf69bd6573a90d7bd287810b5c3746841cee6a7f77cadd016f34a97fc7b2361) with files /root/.config/tac/agent.json.lock and /run/tac-agent.pid; and a check for passwordless doas - A second GreyNoise cluster (GT12, seen 12 Nov 2026) installs a Python implant and the chisel tunnel tool toward 91[.]149[.]254[.]143 port 853 from 140[.]245[.]33[.]114, and persists with an SSH key and fake tiny-cloud services. GreyNoise says some addresses may be compromised legitimate hosts ## Timeline | Date | Event | Source | | --- | --- | --- | | 4 Nov 2026 | Damien Miller writes the OpenSSH fix at 21:18 UTC. The OpenSSH team says it learned of the problem that day. | [source](https://api.github.com/repos/openssh/openssh-portable/commits/4353fdc43ef646c43a8dda7025cc1a898caf1482) | | 5 Nov 2026 | The fix goes public at 09:00 UTC, OpenSSH 10.6p2 is released and announced as an urgent security release, and Alpine publishes its advisory and fixed packages. The reporter is Ozgun Kultekin of Wakefield. The researcher's write-up follows at 12:19 UTC, exploit code appears on GitHub at 13:16 UTC, and the CVE record is published at 15:46 UTC, 2.5 hours after the exploit code. NVD lists it at 16:17 UTC (CVSS 8.1). | [source](https://marc.info/?l=openssh-unix-dev&m=179388064263131&w=2) | | 6 Nov 2026 | Medcrypt publishes an independent reproduction on a stock Yocto image, with mitigation guidance. | [source](https://www.medcrypt.com/blogs/asterisk-hole-cve-2026-123158-when-a-locked-account-is-the-way-in) | | 7 Nov 2026 | GreyNoise observes the first successful exploitation at 00:46 UTC on one of its sensors. Intrusions with follow-up commands begin the same day. | [source](https://sshole.events.greynoise.io/) | | 9 Nov 2026 | CISA adds CVE-2026-123158 to the Known Exploited Vulnerabilities catalog, due 12 Nov 2026. Medcrypt posts its independent reproduction to the oss-security list the same day. | [source](https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json) | | 10 Nov 2026 | NVD last modifies the record (status Analyzed). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-123158) | Exploited 2 days after disclosure: disclosed 5 Nov 2026, first exploitation reported 7 Nov 2026. ## Exploitation reported so far Exploit code was public on GitHub on 5 Nov 2026, the day of the fix, and GreyNoise saw the first successful exploitation on 7 Nov 2026 at 00:46 UTC. CISA added the flaw to KEV on 9 Nov 2026 and lists ransomware use as unknown. Medcrypt reproduced the bypass on a stock Yocto image on 6 Nov 2026. GreyNoise tracks the activity on its event page, updated 19 Nov 2026. Its cluster GT10 used the bypass, checked for passwordless doas, installed a Python runtime and an agent that talks to tac[.]vpnla[.]org, and set up a startup service. A second cluster, GT12, installed a Python implant and a tunnel tool. GreyNoise says not all activity belongs to these clusters, and that it cannot see from the network whether a login used the bypass or a real password. ## Why static reachability settles it Exposure rests on build and configuration facts that a read of the package and config settles: musl, no PAM, password authentication on, and a * account next to a hashed account. Those are static properties of the host, so a scan of the build and /etc/shadow gives the answer before any attack arrives, and a severity score or exploit feed cannot say which hosts meet all of them. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-19868: SonicWall SonicOS SSLVPN, SonicOS SSLVPN SAML bypass](/ledger/cve-2026-19868/) - [CVE-2026-34486: Apache Software Foundation Apache Tomcat](/ledger/cve-2026-34486/) - [CVE-2026-119977: WordPress Core, WordPress core upgrade-window RCE](/ledger/cve-2026-119977/) ## Sources - CVE-2026-123158: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-123158 (read 19 Nov 2026) - CVE-2026-123158: CISA KEV feed (catalog version 2026.11.18): https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - CVE-2026-123158: vendor advisory (OpenSSH announcement 'urgent security release: 10.6p2' on openssh-unix-dev): https://marc.info/?l=openssh-unix-dev&m=179388064263131&w=2 (read 19 Nov 2026) - CVE-2026-123158: vendor advisory (OpenSSH 10.6p2 release notes): https://www.openssh.org/releasenotes.html#10.6p2 (read 19 Nov 2026) - CVE-2026-123158: OpenSSH fix commit (GitHub API): https://api.github.com/repos/openssh/openssh-portable/commits/4353fdc43ef646c43a8dda7025cc1a898caf1482 (read 19 Nov 2026) - CVE-2026-123158: distribution advisory (Alpine Linux security advisory): https://alpinelinux.org/posts/2026-11-05-openssh-password-authentication-bypass/ (read 19 Nov 2026) - CVE-2026-123158: GreyNoise event page and timeline: https://sshole.events.greynoise.io/ (read 19 Nov 2026) - CVE-2026-123158: researcher write-up (Ozgun Kultekin, Wakefield): https://sshhole.com (read 19 Nov 2026) - CVE-2026-123158: Medcrypt independent reproduction and mitigation guidance: https://www.medcrypt.com/blogs/asterisk-hole-cve-2026-123158-when-a-locked-account-is-the-way-in (read 19 Nov 2026) - CVE-2026-123158: oss-security thread post by Medcrypt: http://www.openwall.com/lists/oss-security/2026/11/09/27 (read 19 Nov 2026) - CVE-2026-123158: Alpine Linux releases page (branch end-of-support dates): https://alpinelinux.org/releases/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-123158/ --- # CVE-2026-19868: SonicWall SonicOS SSLVPN, is it exploitable where you run it Ledger entry, disclosed 5 Nov 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-19868 affects SonicWall SonicOS SSLVPN and was disclosed 5 Nov 2026 with a CVSS score of 7.1. Exposure needs these conditions: SonicOS Gen7 runs 7.3.3-7015 or older, or Gen8 runs 8.2.2-8015 or older, and the SSLVPN portal is reachable by the attacker; SAML is on for SSLVPN logins, alongside an active LDAP, RADIUS or TACACS+ server. ## What the flaw is CVE-2026-19868 is an authentication enforcement flaw (CWE-308, use of single-factor authentication) in the SonicOS SSLVPN service. When SAML is used for SSLVPN logins, a user who holds valid credentials for an alternate authentication method can bypass the SAML step. That also skips any multi-factor check the identity provider enforces through SAML. What the attacker gains is an SSLVPN session that the SAML policy was meant to block. They need working credentials for another method configured on the firewall, so this defeats MFA and identity-provider conditional access rather than giving access to a stranger. SonicWall rates it 7.1 (CVSS 3.1 base score; network, high complexity, low privileges, scope changed, high confidentiality impact). NVD carries a CISA-supplied 7.5 CVSS base score that treats privileges as none. SonicWall states the flaw affects the SSLVPN service only if SAML is enabled for SSLVPN logins. It is one of seven CVEs fixed in SonicWall advisory SNWLID-2026-0014, and the same firmware fixes the others. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | SonicOS Gen7 runs 7.3.3-7015 or older, or Gen8 runs 8.2.2-8015 or older, and the SSLVPN portal is reachable by the attacker | Read the firmware version shown in the SonicOS management interface for the firewall and compare it with the versions table. Gen7 needs 7.3.4-7013 or higher, Gen8 needs 8.2.3-8015 or higher. Check every firewall in a pair and every NSv. Match the model against the platform lists in the versions table (Gen7 TZ, NSa, NSsp, NSv; Gen8 TZ, NSa, NSv). Then confirm the SSLVPN portal address answers from the internet or other untrusted networks. | | SAML is on for SSLVPN logins, alongside an active LDAP, RADIUS or TACACS+ server | In the SonicOS management interface, review the SSLVPN user authentication settings and confirm whether SAML is the configured login method for SSLVPN users; SonicWall says the flaw affects SSLVPN only in that case. Then open Device > Users > Settings > Authentication > Configure RADIUS > Settings > RADIUS Servers, then Configure LDAP > Settings > LDAP Servers, then Configure TACACS+ > Settings > TACACS+ Servers. Any server listed and not disabled meets the second half. If either half fails, the bypass does not apply to that firewall. | | the attacker holds valid credentials for an account in that LDAP, RADIUS or TACACS+ directory | List who has accounts in the directory the firewall uses (CVSS privileges required: low); any such account holder can attempt the bypass. | ## Affected and fixed versions | Platforms | Affected versions | Fixed versions | | --- | --- | --- | | Gen7 hardware: TZ 270, 270W, 370, 370W, 470, 470W, 570, 570W, 570P, 670; NSa 2700, 3700, 4700, 5700, 6700; NSsp 10700, 11700, 13700, 15700. Gen7 NSv 270, 470, 870 (ESX, KVM, Hyper-V, AWS, Azure) | 7.3.3-7015 and older versions | 7.3.4-7013 and higher versions | | Gen8 hardware: TZ 80, 280, 280W, 380, 380W, 480, 580, 680; NSa 2800, 3800, 4800, 5800. Gen8 NSv XS, S, M, L (ESX, KVM, Hyper-V, AWS, Azure) | 8.2.2-8015 and older versions | 8.2.3-8015 and higher versions | These are the versions SonicWall lists in SNWLID-2026-0014. NVD's configuration data for this CVE (read 19 Nov 2026) uses different cut-offs: Gen7 before 7.3.4-7015 and Gen8 before 8.2.2-8015. A Gen7 build of 7.3.4-7013 is fixed per SonicWall but falls inside NVD's range, and 8.2.2-8015 is affected per SonicWall but outside NVD's range. Use SonicWall's figures: the Gen7 fix is 7.3.4-7013 and the Gen8 fix is 8.2.3-8015. ## If the fix cannot go in yet - Upgrade to Gen7 7.3.4-7013 or higher, or Gen8 8.2.3-8015 or higher. SonicWall urges customers to upgrade as soon as possible. This is the fix. Cost: a SonicOS firmware upgrade restarts the firewall, which drops active SSLVPN sessions, so schedule a window. The advisory states no downtime figure and gives no order for HA pairs; upgrade both units of a pair. - If SAML is in use and you cannot upgrade, disable or delete every active LDAP, RADIUS and TACACS+ authentication server on the firewall (SonicWall workaround, updated 9 Nov 2026). Paths: Device > Users > Settings > Authentication > Configure RADIUS > Settings > RADIUS Servers, Configure LDAP > Settings > LDAP Servers, and Configure TACACS+ > Settings > TACACS+ Servers; disable or delete the servers in each. Cost: anything on the firewall that signs in through those servers (for example local admin or user logins that use LDAP or RADIUS) stops working until you re-enable them, so confirm what depends on them first. ## What to look for - SonicWall publishes no log message, file path or indicator address that marks a SAML bypass, and the advisory names no SonicOS log event for it. - A check that follows from the flaw, not from a vendor statement: compare SSLVPN login records on the firewall with your identity provider's sign-in log. A user who appears on the firewall with no matching SAML sign-in at the identity provider, or who is listed as authenticated through the LDAP, RADIUS or TACACS+ domain while SAML is required, signed in through the alternate method. ## Timeline | Date | Event | Source | | --- | --- | --- | | 5 Nov 2026 | SonicWall advisory SNWLID-2026-0014 version 1.0, initial release, dated 5 Nov 2026 in the revision history; the PSIRT data stamps publication at 6 Nov 2026 10:40 UTC. | [source](https://psirt.global.sonicwall.com/vuln-detail/SNWLID-2026-0014) | | 6 Nov 2026 | NVD publishes the record at 19:17 UTC. CISA adds the CVE to KEV the same day with a due date of 20 Nov 2026, 14 days later, and its SSVC entry in NVD records exploitation as active. Ransomware use is listed as Unknown. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-19868) | | 9 Nov 2026 | SonicWall advisory version 1.1 updates the workaround for this CVE and the researcher credits. The reporter is credited as Ryan Harris of The Electronic Office. | [source](https://psirt.global.sonicwall.com/vuln-detail/SNWLID-2026-0014) | The earliest dated mark of exploitation is CISA's, on 6 Nov 2026, 1 day after disclosure; exploitation may have begun earlier. ## Exploitation reported so far SonicWall says its PSIRT investigated a report indicating potential exploitation of this vulnerability and urges customers to upgrade. It gives no date for the first attack and no indicators. CISA's SSVC data in NVD records exploitation as active and automatable as no. ## Why static reachability settles it Whether the bypass applies comes down to configuration: SAML on SSLVPN logins plus at least one active LDAP, RADIUS or TACACS+ server, on an affected firmware. Those are facts you can read from the firewall, so a static check of the device settings settles exposure without any traffic analysis. Where all three hold, the exploited status makes the upgrade urgent. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-119977: WordPress Core, WordPress core upgrade-window RCE](/ledger/cve-2026-119977/) - [CVE-2026-123158: OpenBSD OpenSSH, OpenSSH musl no-PAM password bypass](/ledger/cve-2026-123158/) - [CVE-2026-76418: Cisco Secure Firewall ASA and FTD](/ledger/cve-2026-76418/) ## Sources - CVE-2026-19868: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-19868 (read 19 Nov 2026) - CVE-2026-19868: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-19868 (read 19 Nov 2026) - CVE-2026-19868: SonicWall PSIRT advisory SNWLID-2026-0014 (web page): https://psirt.global.sonicwall.com/vuln-detail/SNWLID-2026-0014 (read 19 Nov 2026) - CVE-2026-19868: SonicWall PSIRT advisory SNWLID-2026-0014 (data feed the PSIRT page loads: affected and fixed versions, workaround, revision history): https://psirtapi.global.sonicwall.com/api/v1/vulndetail/?advisory_id=SNWLID-2026-0014 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-19868/ --- # CVE-2026-119977: WordPress Core, is it exploitable where you run it Ledger entry, disclosed 28 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-119977 affects WordPress Core and was disclosed 28 Oct 2026 with a CVSS score of 8.1. Exposure needs these conditions: version: the WordPress core in use is below the first fixed release for its branch; a core upgrade runs on this site: manual, WP-CLI or automatic. CISA added it to the Known Exploited Vulnerabilities catalog on 3 Nov 2026. ## What the flaw is WordPress core unpacks the update package into a working folder under wp-content/upgrade during a full-package core upgrade. In versions before the fix, that folder was named after the package file, so its name could be predicted, and the upgrade folder was served by the web server with no index file. NVD describes it as remote code execution by unauthenticated attackers through predictable, web-accessible staging paths during the upgrade window. The folder naming comes from a comparison of the 7.1.3 and 7.1.4 code. KEV classes it as a time-of-check time-of-use race condition (CWE-367). The window is short and exists only while an upgrade is running, so NVD's CNA score (CVSS 3.1 of 8.1) rates attack complexity High. Sites that accept automatic background updates open the window on their own without an admin present. Because the working folder is created by the updater code already installed, the upgrade that installs the fix still runs the old logic. In version 7.1.4 the folder gets a random 16 character suffix, the upgrade folder gets an index.php so it cannot be listed, and the updater refuses to copy a wp-config.php from the working folder. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | Version: the WordPress core in use is below the first fixed release for its branch. | Run wp core version (WP-CLI), or grep wp_version wp-includes/version.php, or open the dashboard under Dashboard > Updates. Compare with the first fixed release for your branch in the versions table: 7.1.4, 7.0.8, 6.9.11 and so on down to 4.7.39. A fixed backport such as 6.9.11 is not affected even though it is numbered below 7.1.4. | | A core upgrade runs on this site: manual, WP-CLI or automatic. | Check whether automatic core updates are on: Dashboard > Updates states that the site is kept updated automatically, and wp config get WP_AUTO_UPDATE_CORE shows an override in wp-config.php. Look at past runs in the web server log for admin update requests and in wp-content/upgrade for leftover folders. | | The staging area wp-content/upgrade is reachable over HTTP. | From a host outside your network, run curl -sI https://your-site.example/wp-content/upgrade/ and read the status. A 200 or a directory listing means reachable. 403 or 404 means the web server already blocks it, which removes this condition. On a patched site the folder also holds an index.php. | | The unpacked working folder has a predictable name (the unfixed behavior). | During or after an upgrade, run ls -la wp-content/upgrade/. A folder named after the package, such as wordpress-, with no random suffix means the unfixed code ran. The 7.1.4 code adds a 16 character random suffix and an index.php. If wp-content/upgrade contains an index.php, the fixed updater has run on this site. | ## Affected and fixed versions | Branch | Affected | First fixed release | | --- | --- | --- | | 7.1 | 7.1.0 through 7.1.3 | 7.1.4 | | 7.0 | 7.0.0 through 7.0.7 | 7.0.8 | | 6.9 | 6.9.0 through 6.9.10 | 6.9.11 | | 6.8 | 6.8.0 through 6.8.11 | 6.8.12 | | 6.7 | 6.7.0 through 6.7.10 | 6.7.11 | | 6.6 | 6.6.0 through 6.6.10 | 6.6.11 | | 6.5 | 6.5.0 through 6.5.13 | 6.5.14 | | 6.4 | 6.4.0 through 6.4.13 | 6.4.14 | | 6.3 | 6.3.0 through 6.3.13 | 6.3.14 | | 6.2 | 6.2.0 through 6.2.14 | 6.2.15 | | 6.1 | 6.1.0 through 6.1.15 | 6.1.16 | | 6.0 | 6.0.0 through 6.0.17 | 6.0.18 | | 5.9 | 5.9.0 through 5.9.19 | 5.9.20 | | 5.8 | 5.8.0 through 5.8.18 | 5.8.19 | | 5.7 | 5.7.0 through 5.7.20 | 5.7.21 | | 5.6 | 5.6.0 through 5.6.22 | 5.6.23 | | 5.5 | 5.5.0 through 5.5.23 | 5.5.24 | | 5.4 | 5.4.0 through 5.4.24 | 5.4.25 | | 5.3 | 5.3.0 through 5.3.26 | 5.3.27 | | 5.2 | 5.2.0 through 5.2.29 | 5.2.30 | | 5.1 | 5.1.0 through 5.1.27 | 5.1.28 | | 5.0 | 5.0.0 through 5.0.30 | 5.0.31 | | 4.9 | 4.9.0 through 4.9.34 | 4.9.35 | | 4.8 | 4.8.0 through 4.8.33 | 4.8.34 | | 4.7 | 4.7.0 through 4.7.38 | 4.7.39 | Affected ranges come from the CNA record in NVD; first fixed releases are the next release in each branch, all dated 28 Oct 2026 on the wordpress.org releases page. WordPress.org says only the latest release is actively supported, and the backports are a courtesy. Newer releases exist for several branches (7.1.5, 7.0.9, 6.9.12, released 18 Nov 2026). ## If the fix cannot go in yet - Update to the fixed release for your branch, preferably the newest release of your branch (7.1.5 for the 7.1 branch, released 18 Nov 2026). Cost: a short upgrade window, which is the exposure itself. On a public site, put the block from the next item in place first, then run the update. - WordPress.org names no workaround in the 7.1.4 release note. The exposure-reducing step that follows from the code change is to deny web access to wp-content/upgrade. For nginx add location ^~ /wp-content/upgrade/ { deny all; } and for Apache add a Require all denied rule for that directory, then reload the web server. Cost: none for visitors, since the folder only holds update files. The rule is derived from the 7.1.4 change, which adds an index.php to that folder, not from a vendor statement. Before the update begins, the block must already be in place, because the vulnerable updater that installs the fix still creates the predictable folder. - Automatic background updates start the window with nobody watching. The release note says sites that support them begin the 7.1.4 update on their own. With the block in place they are safe to leave on. Holding a site on an old release to avoid the window keeps it exposed to every later upgrade. - After any upgrade on an unpatched site, review the site for unexpected files (see detection). WordPress.org gives no further workaround. ## What to look for - Version check: wp core version, or grep wp_version wp-includes/version.php, compared with the versions table. - After each upgrade on a site that ran code before 7.1.4: list wp-content/upgrade/ and remove stale folders. In the unfixed code the working folder is named after the package, so look for a leftover folder there, or any .php file you did not put there. - Check core files against the official checksums: wp core verify-checksums reports modified and unknown files in the WordPress root and wp-includes. Compare wp-config.php with your last known good copy, because the 7.1.4 updater refuses to copy a wp-config.php from its working folder. Look for new admin users (wp user list --role=administrator) and new files under wp-content/uploads and wp-content/plugins. - In web server logs, look at requests for /wp-content/upgrade/ paths made in the minutes around a core upgrade. A 200 from a path under that folder that is not an update file is suspect. - WordPress.org, Wordfence and CISA publish no indicator addresses or payload strings for this flaw. CISA marks forensic triage as required in KEV. ## Timeline | Date | Event | Source | | --- | --- | --- | | 28 Oct 2026 | WordPress 7.1.4 security release announced. It fixes a weakness in the WordPress upgrade process that can lead to remote code execution during the upgrade window, reported by Umut Bayram, together with a set of cross-site scripting and other fixes. WordPress 7.2 Beta 3 is released with the fixes. Backports for every branch back to 4.7 (7.0.8 down to 4.7.39) are released the same day, as the releases page shows. | [source](https://wordpress.org/news/2026/10/wordpress-7-1-4) | | 31 Oct 2026 | NVD publishes the record from the Wordfence CNA at 06:16 UTC. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-119977) | | 3 Nov 2026 | CISA adds the CVE to KEV (time-of-check time-of-use race condition), due date 6 Nov 2026, forensic triage Yes. CISA's SSVC entry in NVD records exploitation as active. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-119977) | | 4 Nov 2026 | NVD last modified the record (status Analyzed). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-119977) | | 18 Nov 2026 | WordPress 7.1.5 and matching releases for older branches are published. | [source](https://wordpress.org/download/releases/) | The earliest dated mark of exploitation is CISA's, on 3 Nov 2026, 6 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far CISA added the CVE to KEV on 3 Nov 2026, three days after NVD published it and six days after the fix, and its SSVC entry in NVD records exploitation as active and not automatable. KEV lists ransomware campaign use as Unknown. CISA publishes no first exploitation date, attacker or indicators for this CVE. The 3 Nov 2026 KEV listing is the earliest dated public report of exploitation, so the days from fix to exploitation on this page measure the gap to the listing, not an observed attack. ## Why runtime reachability settles it Exposure needs an upgrade running at that moment and an upgrade folder the attacker can reach, which are runtime conditions. Checking whether the staging path answers over HTTP and whether automatic updates are on tells you more than the version number does. The fixed version still ends the question. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-76418: Cisco Secure Firewall ASA and FTD](/ledger/cve-2026-76418/) - [CVE-2026-19868: SonicWall SonicOS SSLVPN, SonicOS SSLVPN SAML bypass](/ledger/cve-2026-19868/) - [CVE-2026-109618: PHP Group PHP](/ledger/cve-2026-109618/) ## Sources - CVE-2026-119977: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-119977 (read 19 Nov 2026) - CVE-2026-119977: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-119977 (read 19 Nov 2026) - CVE-2026-119977: vendor advisory (WordPress 7.1.4 security release note): https://wordpress.org/news/2026/10/wordpress-7-1-4 (read 19 Nov 2026) - CVE-2026-119977: WordPress HelpHub page for version 7.1.4: https://wordpress.org/documentation/wordpress-version/version-7-1-4/ (read 19 Nov 2026) - CVE-2026-119977: WordPress.org releases page (release dates of backports): https://wordpress.org/download/releases/ (read 19 Nov 2026) - CVE-2026-119977: WordPress core source, 7.1.3 tag, class-wp-upgrader.php (before the fix): https://raw.githubusercontent.com/WordPress/wordpress-develop/7.1.3/src/wp-admin/includes/class-wp-upgrader.php (read 19 Nov 2026) - CVE-2026-119977: WordPress core source, 7.1.4 tag, class-wp-upgrader.php (after the fix): https://raw.githubusercontent.com/WordPress/wordpress-develop/7.1.4/src/wp-admin/includes/class-wp-upgrader.php (read 19 Nov 2026) - CVE-2026-119977: WordPress core source, 7.1.3 tag, update-core.php: https://raw.githubusercontent.com/WordPress/wordpress-develop/7.1.3/src/wp-admin/includes/update-core.php (read 19 Nov 2026) - CVE-2026-119977: WordPress core source, 7.1.4 tag, update-core.php: https://raw.githubusercontent.com/WordPress/wordpress-develop/7.1.4/src/wp-admin/includes/update-core.php (read 19 Nov 2026) - CVE-2026-119977: WordPress.org version check API (current releases per branch): https://api.wordpress.org/core/version-check/1.7/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-119977/ --- # CVE-2026-76418: Cisco Secure Firewall ASA and FTD, is it exploitable where you run it Ledger entry, disclosed 28 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-76418 affects Cisco Secure Firewall ASA and FTD and was disclosed 28 Oct 2026 with a CVSS score of 8.6. Exposure needs these conditions: the release is earlier than the first fixed release for its train, with Remote Access SSL VPN enabled on an interface the attacker can reach; a tunnel group in scope uses secondary authentication. ## What the flaw is CVE-2026-76418 is a flaw in the Remote Access SSL VPN service of Cisco Secure Firewall ASA Software and Secure Firewall Threat Defense (FTD) Software. Cisco attributes it to logic errors when secondary authentication is configured on a tunnel group; CISA labels the weakness an unchecked return value leading to a NULL pointer dereference (CWE-690). An unauthenticated attacker on the network can send a crafted request to the SSL VPN login service and make the device reload unexpectedly. The result is denial of service: VPN users lose their sessions and the firewall is offline until it restarts. Cisco says a successful VPN login is not needed to reach the affected code. Cisco rates it 8.6 (CVSS 3.1 base score; scope changed, availability impact high, no confidentiality or integrity impact). ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the release is earlier than the first fixed release for its train, with Remote Access SSL VPN enabled on an interface the attacker can reach | Version: on ASA run `show version` and read the Software Version. On FTD run `show version` in the FTD CLI, or open Devices > Device Management in Secure FMC (or Device > Summary in FDM). Compare with the versions table, or enter the release in the Cisco Software Checker. SSL VPN enabled: on ASA run `show running-config webvpn` and look for `enable `. On FTD, remote access VPN is enabled from Devices > VPN > Remote Access in Secure FMC or from Remote Access VPN in FDM, so review the deployed policy there and the resulting running configuration. Reach: take the interface named in the `enable` line, check its IP address and the access rules or upstream firewall that expose it, and test the login page from an untrusted network. An internet-facing interface is the high-risk case. | | a tunnel group in scope uses secondary authentication | Run `show running-config tunnel-group` and search for `secondary-authentication-server-group` in every tunnel group, including DefaultWEBVPNGroup. A group is in scope when `tunnel-group-list enable` is set (check all groups, not only those shown on the login page) or, when it is not set or was removed with `no tunnel-group-list enable`, when it is the default group DefaultWEBVPNGroup; disabling the list does not remove exposure when that default group uses secondary authentication. | ## Affected and fixed versions | Product and train | Affected | First fixed release | | --- | --- | --- | | ASA 9.16 and earlier | Earlier than 9.16.4.104 | 9.16.4.104 | | ASA 9.18 | Earlier than 9.18.4.96 | 9.18.4.96 | | ASA 9.20 | Earlier than 9.20.4.50 | 9.20.4.50 | | ASA 9.22 | Earlier than 9.22.3.200 | 9.22.3.200 | | ASA 9.23 | Earlier than 9.23.1.48 | 9.23.1.48 | | ASA 9.24 | Earlier than the fixed release for the running minor train | 9.24.1.27 and 9.24.10.1 (Cisco lists both in the 9.24 row without naming the train each serves; the Software Checker returns the one for your running release) | | FTD 7.0 and earlier | Earlier than 7.0.11 | 7.0.11 | | FTD 7.2 | Earlier than 7.2.13 | 7.2.13 | | FTD 7.4 | Earlier than 7.4.9 | 7.4.9 | | FTD 7.6 | Earlier than 7.6.7 | 7.6.7 | | FTD 7.7 | Earlier than 7.7.14 | 7.7.14 | | FTD 10.0 | Earlier than 10.0.3 | 10.0.3 | | FTD 10.1 | Earlier than 10.1.1 | 10.1.1 | Cisco confirms these products are not affected: Cloud-delivered Firewall, cdFMC, Secure Access and Secure FMC Software. The Cisco Software Checker returns the first fixed release for a given running release. Cisco validates only the affected and fixed release information in the advisory. ## If the fix cannot go in yet - Upgrade to the first fixed release for your train. Cisco strongly recommends this and says no workaround addresses the flaw; the two cases below are a temporary mitigation only. - Mitigation, Case 1: if DefaultWEBVPNGroup does not use secondary authentication, disable the tunnel group list with `no tunnel-group-list enable` in WebVPN configuration mode. Cost: users can no longer pick a connection profile from a list on the login page and must know the right profile name. - Mitigation, Case 2: if DefaultWEBVPNGroup uses secondary authentication, remove it with `no secondary-authentication-server-group`. Cost: the remote access VPN security posture drops, since users are no longer asked for the second credential. Cisco says the mitigation was proven in a test environment, tells customers to check it in their own environment first, and prefers the upgrade. - The aim is that no tunnel group in scope uses secondary authentication. In Case 1 the group list is off, so the only group in scope is DefaultWEBVPNGroup, which has none. In Case 2 secondary authentication is removed from DefaultWEBVPNGroup, and any user-configured group in an enabled group list must lose it too. ## What to look for - After a suspected attack, `show crashinfo` on the ASA or FTD lists the crash or traceback record that an unexpected reload leaves, and `show version` shows the uptime since the last reload. A reload with no planned change, repeated while the SSL VPN is reachable, matches the symptom. - Cisco publishes no log strings, syslog message IDs or indicator addresses for this flaw. For FTD, collect a troubleshoot file from Secure FMC or FDM before sharing crash data with Cisco TAC. ## Timeline | Date | Event | Source | | --- | --- | --- | | 28 Oct 2026 | Cisco publishes advisory cisco-sa-asaftd-tunnel-dos-s6mgkb75 version 1.0 (Final) with fixed releases (bug ID CSCwx26625); the finder is credited as Eric Fernald through a Cisco TAC support case. NVD publishes the record the same day, and CISA's SSVC data in it marks exploitation as active. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-76418) | | 29 Oct 2026 | CISA adds the CVE to KEV with a due date of 1 Nov 2026, three days after the add. Ransomware use is listed as Unknown. The KEV required action is to apply mitigations in accordance with vendor instructions, follow CISA's Forensics Triage Requirements and discontinue use of the product if mitigations are unavailable. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-76418) | The earliest dated mark of exploitation is CISA's, on 28 Oct 2026, the day of disclosure; exploitation may have begun earlier. ## Exploitation reported so far Cisco states that in October 2026 its PSIRT became aware of active exploitation of this vulnerability, without naming a day, and recommends upgrading. The earliest dated report is CISA's SSVC entry in NVD, stamped 28 Oct 2026, the day of the advisory, with exploitation active and automatable yes. CISA added the CVE to KEV on 29 Oct 2026. No attacker names, campaign details or indicators are published. KEV lists ransomware campaign use as Unknown. ## Why static reachability settles it Exposure needs a vulnerable release, SSL VPN enabled on an interface and secondary authentication in a tunnel group in scope. The version and both settings show in the running configuration, so a static check of the config settles those three exactly. Whether the login service is reachable depends on the network path, so test it from an untrusted network. A device that matches all of these is exposed to a no-login attack that is already being used. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-109618: PHP Group PHP](/ledger/cve-2026-109618/) - [CVE-2026-119977: WordPress Core, WordPress core upgrade-window RCE](/ledger/cve-2026-119977/) - [CVE-2026-109619: PHP Group PHP](/ledger/cve-2026-109619/) ## Sources - CVE-2026-76418: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-76418 (read 19 Nov 2026) - CVE-2026-76418: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-76418 (read 19 Nov 2026) - CVE-2026-76418: Cisco security advisory cisco-sa-asaftd-tunnel-dos-s6mgkb75 (vulnerable configuration, mitigation, fixed releases): https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-asaftd-tunnel-dos-s6mgkb75 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-76418/ --- # CVE-2026-109618: PHP Group PHP, is it exploitable where you run it Ledger entry, disclosed 21 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-109618 affects PHP Group PHP and was disclosed 21 Oct 2026 with a CVSS score of 9.8. Exposure needs these conditions: version: the PHP build serving web requests is below 8.2.35, 8.3.36, 8.4.27 or 8.5.12; a web SAPI is in use: mod_php, PHP-FPM, php-cgi, FrankenPHP or LiteSpeed. CISA added it to the Known Exploited Vulnerabilities catalog on 26 Oct 2026. ## What the flaw is PHP leaves a request through longjmp() when it runs out of memory. In some places during request startup, request memory is reachable before it is fully set up. If any allocation fails at that point, PHP skips the cleanup that would reset it, and the request shutdown code later reads or branches on the invalid memory. The failed allocation does not need to be related to the broken memory. CWE-416. The confirmed result is unauthenticated remote code execution as the web worker account. PHP's advisory says this was validated over HTTP against Apache mod_php with default input and memory limits and normal prefork configurations. Extra workers and unrelated traffic raise the number of attempts needed but do not remove the condition. Several multipart parsing problems let a request reach the out-of-memory state while staying inside its apparent PHP limits. Reliability depends on the PHP build, extensions, frontend and allocator layout. Ratings: NVD's primary CVSS 3.1 base score is 9.8 (AV:N/AC:L/PR:N/UI:N). PHP's own CVSS 4.0 base score, also in NVD, is 9.2 with attack complexity High. CVE-2026-109619 is the sibling flaw from the same release, and PHP says to follow the remediation advice for both. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | Version: the PHP build serving web requests is below 8.2.35, 8.3.36, 8.4.27 or 8.5.12. | Run the binary of the web SAPI, not only the CLI: php-fpm -v, php-cgi -v, or for mod_php run php -v and confirm with apachectl -M \| grep -i php (Debian and Ubuntu: apache2ctl -M \| grep -i php). For packages run rpm -q php or dpkg -l 'php*'. Packages may be patched without a version change, so also run rpm -q --changelog php \| grep -i 109618 or apt changelog php8.3 \| grep -i 109618. Check every host and every container image, including PHP bundled inside appliances and Docker base images. | | A web SAPI is in use: mod_php, PHP-FPM, php-cgi, FrankenPHP or LiteSpeed. | Run ps -ef \| grep -Ei 'php-fpm\|php-cgi\|frankenphp\|lsphp\|httpd\|apache2' and apachectl -M \| grep -i php. If PHP runs only from the command line for cron jobs, this condition is false for that host. | | An unauthenticated client can reach a PHP script and send it a multipart POST. | From a host outside your network, send a POST with no credentials to a known PHP entry point such as /index.php and see whether the server hands it to PHP (any PHP response, even an error page, means yes). Review which vhosts and locations pass .php to PHP in the web server configuration (grep -R 'fastcgi_pass\\|SetHandler\\|proxy:fcgi\\|AddHandler' /etc/nginx /etc/apache2 /etc/httpd). A frontend that blocks multipart bodies, or that requires login for every PHP path, removes this condition. | | A finite memory_limit applies to the worker that handles the request. | Look at the value each SAPI uses, not only the CLI: grep -Rn 'memory_limit' /etc/php* /usr/local/etc/php* /etc/php-fpm.d /etc/apache2 /etc/httpd. Check pool files for php_admin_value[memory_limit] and php_value[memory_limit], Apache php_value / php_admin_value lines, and .htaccess files. A value of -1 means no limit, which reduces the confirmed exploit but does not make the host safe. | ## Affected and fixed versions | PHP branch | Affected | First fixed release | | --- | --- | --- | | PHP 8.5 | Earlier than 8.5.12 | 8.5.12 | | PHP 8.4 | Earlier than 8.4.27 | 8.4.27 | | PHP 8.3 | Earlier than 8.3.36 | 8.3.36 | | PHP 8.2 | Earlier than 8.2.35 | 8.2.35 | | PHP 8.1 and older | Unsupported branches; PHP publishes no fix | None upstream | GitHub's advisory and NVD list the four supported branches. CISA tells operators still on PHP 8.1 or older to take PHP-based systems offline until they are patched or updated, and Fastly tells them to ask their package distributor about backports or upgrade. Distribution packages often keep the old version number and add the fix as a backport, so check the package changelog for the CVE ID as well as the version. ## If the fix cannot go in yet - Upgrade to 8.5.12, 8.4.27, 8.3.36 or 8.2.35, or to your distribution's patched package. PHP calls this the only complete mitigation. Cost: a PHP restart, and the usual regression testing for a minor release. The patch replaces longjmp() on out-of-memory with a delayed soft failure at the next VM safe point and an immediate hard exit through _exit() at a second, higher memory threshold. - If you cannot update: set memory_limit=-1 administratively and restart every PHP worker, then confirm no pool, directory or application setting restores a finite limit. Cost: a runaway script is no longer stopped by PHP and can use all host memory. This is not complete, since a failed mmap() can still hit the vulnerable path. - Keep post_max_size conservative and enforce the decoded request size at the frontend. Lower max_input_nesting_level, and bound input variables, multipart parts, headers per part and uploads. Cost: large uploads and deeply nested forms are refused. PHP says these controls reduce out-of-memory and denial of service risk even after patching. - If the application does not use multipart form data, block multipart requests at the frontend before they reach PHP. Cost: file uploads and multipart forms stop working. - A container or cgroup memory limit (memory.max) can stand in for PHP's own limit, with caveats. It does not guarantee allocation failure is avoided, and killing a whole container or cgroup can kill the FPM master and all workers unless each worker has its own group. Do not use RLIMIT_AS or RLIMIT_DATA as a substitute: they make mmap() fail and trigger the vulnerable path. - CISA's guidance for systems that cannot be updated, and for PHP 8.1 and older, is to take them offline until patched. A WAF may give partial protection in the meantime. Fastly's Next-Gen WAF has a CVE-specific rule covering both PHP2Shell CVEs; opt in to it. Cost of blocking IP lists: Fastly warns that several circulating lists include VPN exits, Tor exits and residential addresses carrying ordinary traffic. ## What to look for - Network indicator (Fastly): the most widespread exploit kit sends POST to /index.php or / with User-Agent php2shell-109618 and Content-Type multipart/form-data with the fixed boundary B7kQpoc618, a fixed-size body and a minimal header set. It targets raw IP addresses and sends one or two requests per host. Fastly says the multipart boundary string is the most reliable separator; User-Agent strings, source IPs and body sizes go stale within a day and mislead. - Rule out look-alikes first (Fastly): chunked multipart requests with no Content-Length, and bodies padded with long runs of the letter A, were 98.6% and 99.9% React2Shell (CVE-2025-55182) traffic in the hour Fastly measured. React2Shell requests carry a Next-Action header, which PHP traffic never does. A WordPress kit with a boundary starting ----wp2shell is a separate July 2026 campaign. - Host evidence: success is not visible at the edge, so check the origin. Look in PHP-FPM error logs and Apache error logs for worker crashes, for example AH00051 'child pid ... exit signal' entries. Look for unexpected outbound connections from PHP workers, since a connect-back shell is origin egress (Fastly). CISA: monitor for PHP worker processes launching shells or downloading tools. - CISA published a SIGMA detection rule for suspicious activity in its alert, updated 23 Oct 2026. Fastly saw no payload string to match, because the exploit is a failure condition. About a third of the PHP2Shell-named traffic Fastly saw labeled itself as authorized exposure or patch-state checking, so scanning volume alone does not equal exploitation. ## Timeline | Date | Event | Source | | --- | --- | --- | | 20 Oct 2026 | The CVE IDs circulate on forums and social media from about 15:00 UTC, with a wrong release time and a wrong CVSS rating. PHP publishes a pre-announcement at 19:30 UTC and moves the coordinated release forward a day to 21 Oct 14:00 UTC (as described by Fastly from the PHP Foundation's postmortem). | [source](https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell) | | 21 Oct 2026 | 03:59 UTC: first request from a PHP2Shell-named tool in Fastly's telemetry. 05:11 UTC: first request from the most widespread exploit kit. 07:41 UTC: Chainguard publishes images containing the fix. 13:34 UTC: GitHub advisories published for both CVEs. 14:00 UTC: coordinated release of PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35, by which time Fastly had seen more than 100 source IPs using PHP2Shell-named tools. 14:17 UTC: NVD publishes the record. CISA publishes an alert urging immediate patching of both PHP CVEs (revised 23 and 26 Oct). | [source](https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell) | | 23 Oct 2026 | CISA updates its alert with a SIGMA detection rule, partner resources and reinforced mitigation guidance. | [source](https://www.cisa.gov/news-events/alerts/2026/10/21/critical-vulnerabilities-php-require-immediate-patching) | | 26 Oct 2026 | CISA adds the CVE to KEV (name: PHP Memory Corruption Vulnerability), due date 29 Oct 2026, forensic triage marked Yes, ransomware use Unknown. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-109618) | | 18 Nov 2026 | NVD last modified the record (status Analyzed). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-109618) | Exploited on the day of disclosure, 21 Oct 2026. ## Exploitation reported so far Exploitation is public and mass-scale. Fastly reports that tools named after the bug appeared in its WAF telemetry from 21 Oct 2026 at 03:59 UTC, more than nine hours before the coordinated release, and that an exploit kit followed at 05:11 UTC. From the first tool through midday on 22 Oct it stored about 250,000 requests from about 3,200 public IPs. Roughly a third labeled themselves as defensive checks. Fastly counts are lower bounds and cannot show which requests succeeded. CISA added the CVE to KEV on 26 Oct 2026 and marks forensic triage as required. KEV lists ransomware campaign use as Unknown. The early traffic came before the fixed releases shipped. ## Why attack-path analysis settles it Exposure needs a chain: an unpatched build, a web SAPI, an unauthenticated route to PHP, and a finite memory limit. Tracing that path on each host, rather than reading the 9.8 base score, shows which servers can be hit. Where the path is open, the version check ends the question. ## Related pages - [CVE-2026-109619: PHP Group PHP](/ledger/cve-2026-109619/) - [CVE-2026-76418: Cisco Secure Firewall ASA and FTD](/ledger/cve-2026-76418/) - [CVE-2026-111497: Node.js Brotli decoder](/ledger/cve-2026-111497/) ## Sources - CVE-2026-109618: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-109618 (read 19 Nov 2026) - CVE-2026-109618: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-109618 (read 19 Nov 2026) - CVE-2026-109618: vendor advisory (PHP GitHub security advisory GHSA-w8hw-w9fx-j79w): https://github.com/php/php-src/security/advisories/GHSA-w8hw-w9fx-j79w (read 19 Nov 2026) - CVE-2026-109618: Fastly report on PHP2Shell exploitation, disclosure timeline and indicators: https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell (read 19 Nov 2026) - CVE-2026-109618: CISA alert: Critical Vulnerabilities in PHP Require Immediate Patching: https://www.cisa.gov/news-events/alerts/2026/10/21/critical-vulnerabilities-php-require-immediate-patching (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-109618/ --- # CVE-2026-109619: PHP Group PHP, is it exploitable where you run it Ledger entry, disclosed 21 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-109619 affects PHP Group PHP and was disclosed 21 Oct 2026 with a CVSS score of 8.1. Exposure needs these conditions: the PHP build serving web requests is below 8.2.35, 8.3.36, 8.4.27 or 8.5.12; an unauthenticated client can reach a PHP worker and trigger a producer event. CISA added it to the Known Exploited Vulnerabilities catalog on 26 Oct 2026. ## What the flaw is A PHP worker keeps global state, such as error, session, object and extension state, that points into memory owned by one request. That state must be reset before the request memory is freed. In some paths it is not reset, or it is filled in again after the reset. The request heap is then freed while the worker still refers to it. A later request handled by the same worker can read, free or branch on that stale state before PHP has initialized the new request, which is a use after free (CWE-416). A prepared heap layout can turn this into remote code execution. Exploitation needs two parts. A producer is a PHP operation in one request that leaves request-owned state published, such as a fatal error, uncaught error, failed require, Maximum execution time exceeded, or out-of-memory failure. A consumer is an operation in a later request that runs before normal initialization and acts on the stale state. PHP confirms two consumers: an INI warning or deprecation raised during request startup, and session upload progress handling. In the demonstrated attacks the two parts are in different HTTP requests on the same worker. The confirmed remote code execution was shown on Apache mod_php, using the PHP 8.5 fatal error backtrace producer. The session upload progress path was validated separately on PHP-FPM. The minimum set for exposure is an affected build, a worker an unauthenticated client can reach, one producer and one consumer. Session upload progress is on by default, so on a default configuration the consumer is already present and only a producer is needed. Ratings: NVD's primary CVSS 3.1 base score is 8.1 (attack complexity High). PHP's CVSS 4.0 base score in NVD is 9.2. PHP says exact affected versions, configuration and mitigations depend on which internal state is involved, and that no configuration setting covers every case. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the PHP build serving web requests is below 8.2.35, 8.3.36, 8.4.27 or 8.5.12 | Run php-fpm -v, php-cgi -v, or the mod_php build's php -v (confirm mod_php with apachectl -M \| grep -i php). For packages: rpm -q php or dpkg -l 'php*', then rpm -q --changelog php \| grep -i 109619 or apt changelog php8.3 \| grep -i 109619, since distributions backport fixes without changing the version. | | an unauthenticated client can reach a PHP worker and trigger a producer event | Run ps -ef \| grep -Ei 'php-fpm\|frankenphp\|lsphp\|httpd\|apache2' and apachectl -M \| grep -i php. Check pm.max_requests in the pool files (grep -R pm.max_requests /etc/php* /etc/php-fpm.d): a low value shortens how long stale state can live but does not remove the flaw. From outside your network, confirm that PHP entry points answer without login. Review which routes, content types, uploads and plugin options can raise a fatal error or time out. PHP says to include upload, import, error page, shutdown, destructor, serialization and output handler paths, not only ordinary page requests. | | session upload progress is on (the default), which supplies a consumer | Read the value the web SAPI uses: run php-fpm -i \| grep session.upload_progress.enabled for PHP-FPM or php-cgi -i for CGI, and for mod_php read phpinfo() output served through Apache. The CLI's php -i reads the CLI configuration and can differ, so it does not answer this check. Then run grep -Rn upload_progress /etc/php* /etc/php-fpm.d /etc/apache2 /etc/httpd. A value of 1 or On means the consumer is present. | | on PHP 8.5, fatal_error_backtraces is enabled, the strongest confirmed producer | Read the web SAPI value, not the CLI's: php-fpm -i \| grep fatal_error_backtraces for PHP-FPM, or phpinfo() output served through Apache for mod_php, plus grep -Rn fatal_error_backtraces /etc/php* /etc/php-fpm.d /etc/apache2 /etc/httpd. Only PHP 8.5 applies. | | INI values that raise a warning or deprecation at request startup, which supply a consumer | Audit startup configuration, then run the logging audit described under What to look for. PHP lists likely sources: auto_detect_line_endings=1, a non-default filter.default, legacy assert.* settings on 8.3 and later, legacy session settings on 8.4 and later, report_memleaks=0 or a nonzero intl.error_level on 8.5, configured mbstring or iconv encodings, invalid or conflicting zend.assertions, memory_limit and date.timezone values, and log, mail, session or SOAP cache paths rejected by open_basedir. Also check Apache php_value / php_flag / php_admin_* lines and .htaccess files, and PHP_VALUE parameters sent by nginx or mod_proxy_fcgi. Pool values and .user.ini are applied at worker start or after initialization and are not early consumers. | ## Affected and fixed versions | PHP branch | Affected | First fixed release | | --- | --- | --- | | PHP 8.5 | Earlier than 8.5.12 | 8.5.12 | | PHP 8.4 | Earlier than 8.4.27 | 8.4.27 | | PHP 8.3 | Earlier than 8.3.36 | 8.3.36 | | PHP 8.2 | Earlier than 8.2.35 | 8.2.35 | | PHP 8.1 and older | Unsupported branches; PHP publishes no fix | None upstream | Same branches and fixed releases as the sibling CVE-2026-109618. Distributions backport fixes without changing the version, so check the package changelog for CVE-2026-109619 as well as the version. ## If the fix cannot go in yet - Upgrade to 8.5.12, 8.4.27, 8.3.36 or 8.2.35, or to your distribution's patched package. PHP says only a release with the complete state-reset patch is reliable, because no generic configuration setting covers every producer and consumer. Cost: a PHP restart and regression testing. - PHP 8.5 only: set fatal_error_backtraces=0 administratively and restart every worker. This removes the producer used by the confirmed RCE case. Cost: fatal error messages no longer carry backtraces, which makes debugging harder. It does not cover other request-owned state. - Set session.upload_progress.enabled=0 administratively and restart every worker. This stops the demonstrated early session producing and consuming operations. Cost: upload progress tracking stops working for any application that uses it. PHP suggests keeping it off after patching if you do not need it. It does not cover the other lifecycle paths. - Remove deprecated and invalid INI directives so no diagnostics fire during request initialization. Prefer php.ini, FPM pool or server configuration applied at startup over per-request overrides (Apache php_value, .htaccess, frontend PHP_VALUE). Changing error_reporting, using @, or turning off display_errors or log_errors does not help. Cost: configuration cleanup time. - If performance does not matter, set pm.max_requests=1 so each worker handles one request. Cost: every request pays worker startup time, and throughput drops sharply. Separately, stop attacker-controlled requests from reaching Maximum execution time exceeded inside an affected operation; changing the time limit only removes that one interruption. - Also apply the mitigations for CVE-2026-109618, since PHP says to follow the remediation advice from the related report. Both are in the same patch release. CISA's guidance for systems that cannot be patched is to take them offline. ## What to look for - PHP's logging audit: with log_errors=On, a functional error_log and error_reporting=E_ALL set in startup configuration, review Apache, PHP-FPM, reverse proxy and service logs for warnings, deprecations and core errors repeated before target scripts run; warnings produced during extension or server cleanup; fatal, parse and uncaught error messages; the string Maximum execution time exceeded; and failures tied to routes, content types, files, callbacks or plugin options an attacker controls. - PHP warns that a clean log is not proof of safety: historical logs may exclude deprecations, an E_ALL audit covers only the routes exercised, and a failure may happen before logging is set up. - Host evidence of compromise (CISA, Fastly): PHP worker processes launching shells or downloading tools, worker crashes in the Apache error log such as AH00051 'child pid ... exit signal', and unexpected outbound connections from PHP workers. - Fastly's PHP2Shell-named tooling covered both CVEs and was not split by CVE, so the network indicators in the CVE-2026-109618 entry (boundary B7kQpoc618 with User-Agent php2shell-109618) are the most specific published signature for both. CISA published a SIGMA rule in its alert. ## Timeline | Date | Event | Source | | --- | --- | --- | | 20 Oct 2026 | The CVE IDs circulate before release. PHP pre-announces at 19:30 UTC and moves the coordinated release to 21 Oct 14:00 UTC (as described by Fastly). | [source](https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell) | | 21 Oct 2026 | First PHP2Shell-named tool traffic in Fastly telemetry at 03:59 UTC; the tooling covered this CVE and CVE-2026-109618 and is not split by CVE. GitHub advisories for both CVEs published at 13:34 UTC. Coordinated release of PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35 at 14:00 UTC. 14:17 UTC: NVD publishes the record. | [source](https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell) | | 23 Oct 2026 | CISA updates its alert with a SIGMA rule and reinforced mitigation guidance. | [source](https://www.cisa.gov/news-events/alerts/2026/10/21/critical-vulnerabilities-php-require-immediate-patching) | | 26 Oct 2026 | CISA adds the CVE to KEV (name: PHP Stale Internal Request-State Vulnerability), due date 29 Oct 2026, forensic triage marked Yes, ransomware use Unknown. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-109619) | | 18 Nov 2026 | NVD last modified the record (status Analyzed). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-109619) | The earliest dated mark of exploitation is CISA's, on 26 Oct 2026, 5 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far PHP says the strongest demonstrated case is a PHP 8.5 fatal error backtrace producer with a consumer on a later request. Success cannot be told from failure at the network edge, so confirmation is on the origin host. This entry dates exploitation to CISA's KEV listing on 26 Oct 2026, 5 days after disclosure, because that is the earliest mark specific to this CVE. Fastly's 21 Oct 2026 traffic came from PHP2Shell-named tools that covered this CVE and CVE-2026-109618 together, and Fastly does not split its counts by CVE, so that date cannot be assigned to this CVE alone. ## Why attack-path analysis settles it Exploitation needs a producer event in one request and a consumer in a later request on the same worker. Whether both exist depends on the PHP branch, the SAPI, the INI settings and the application's error paths, so tracing that path on each deployment settles exposure. Where the build is fixed, the path no longer matters. ## Related pages - [CVE-2026-111497: Node.js Brotli decoder](/ledger/cve-2026-111497/) - [CVE-2026-109618: PHP Group PHP](/ledger/cve-2026-109618/) - [CVE-2026-102764: BeyondTrust Remote Support](/ledger/cve-2026-102764/) ## Sources - CVE-2026-109619: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-109619 (read 19 Nov 2026) - CVE-2026-109619: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-109619 (read 19 Nov 2026) - CVE-2026-109619: vendor advisory (PHP GitHub security advisory GHSA-6wmf-4rv5-p3rp): https://github.com/php/php-src/security/advisories/GHSA-6wmf-4rv5-p3rp (read 19 Nov 2026) - CVE-2026-109619: Fastly report on PHP2Shell exploitation, disclosure timeline and indicators: https://www.fastly.com/blog/how-fastly-detected-and-built-a-coordinated-response-to-php2shell (read 19 Nov 2026) - CVE-2026-109619: CISA alert: Critical Vulnerabilities in PHP Require Immediate Patching: https://www.cisa.gov/news-events/alerts/2026/10/21/critical-vulnerabilities-php-require-immediate-patching (read 19 Nov 2026) - CVE-2026-109619: vendor advisory for the sibling CVE-2026-109618 (GHSA-w8hw-w9fx-j79w): https://github.com/php/php-src/security/advisories/GHSA-w8hw-w9fx-j79w (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-109619/ --- # CVE-2026-111497: Node.js Brotli decoder, is it exploitable where you run it Ledger entry, disclosed 21 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-111497 affects Node.js Brotli decoder and was disclosed 21 Oct 2026 with a CVSS score of 9.8. Exposure needs these conditions: the Node.js release in use is below the fixed release for its line, or another product embeds a Brotli copy older than 1.0.10, 1.1.1 or 1.2.1; an attacker can feed it Brotli-compressed input over the network. ## What the flaw is Brotli splits compressed data into blocks, and a block switch says which block type comes next. When a stream declares only one block type for a kind, a switch is not valid and the decoder should reject the stream. The block-switch helper returned a plain boolean, so it gave the same answer for an invalid switch as for a stream that simply ran out of input. The caller was told to send more input. It did, and the decoder kept the extra bytes in an eight-byte internal buffer that never drained. Past the eighth byte the writes land outside the buffer (CWE-252 unchecked return value, leading to CWE-787 out-of-bounds write). An attacker who can get Brotli data in front of a vulnerable decoder can corrupt process memory. Node.js lists a crash, disclosure of process memory and remote code execution as the outcomes. The bug shipped in Brotli 1.0.3 in March 2018, and the fix was committed upstream on 27 Feb 2026 and released on 21 Oct 2026. Node.js published its advisory and fixed builds on 21 Oct 2026 (the disclosure date on this page); NVD published the record a day later, on 22 Oct 2026. The flaw is in Google's Brotli, so it affects far more than Node.js: any product that embeds the decoder is exposed. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the Node.js release in use is below the fixed release for its line, or another product embeds a Brotli copy older than 1.0.10, 1.1.1 or 1.2.1 | Node.js: the fixed releases are in the versions table; end-of-life lines (20.x, 23.x, 25.x and earlier) have no fix. Run `node --version` on every host, CI runner and developer machine. For container images run `docker run --rm node --version`. Do not use `process.versions.brotli` as a test: Node.js cherry-picked the fix without changing the bundled version string, so the value is the same on patched and vulnerable builds of a line (ELLIO read 1.2.0 on builds of both kinds; a 22.x build prints 1.1.0). Compare the Node.js release number instead. A Dockerfile pinned to node:24.22.0 or node:22.23.3 stays vulnerable until the pin moves. Other copies of Brotli: on Linux run `ldd $(command -v node) \| grep -i brotli`; a hit means Node.js links the system library. Then check the package: `dpkg -l \| grep -i brotli` (Debian, Ubuntu), `rpm -q brotli` (RHEL, Fedora) or `apk info -v \| grep brotli` (Alpine). Debian's tracker records that its nodejs package uses the system Brotli. ELLIO lists fixed distribution packages: Debian bookworm 1.0.9-2+deb12u1, Debian trixie 1.1.0-2+deb13u1, Alpine 1.2.1-r0 (edge, 3.24, 3.23) and 1.1.1-r0 (3.21). For other products, run an SBOM or composition scan for brotli, libbrotli and the language bindings. Examples with published fixed versions (ELLIO): Bun below 1.4.4 (`bun --version`), PyPI Brotli below 1.2.1 and brotlicffi below 1.2.1.0 (`pip show brotli brotlicffi`), Envoy where the Brotli decompressor is configured (fixed in 1.40.0, 1.39.3, 1.38.6, 1.37.8, 1.36.12), Ruby brotli gem (0.9.1, 0.9.2). NVD also lists Siemens products as affected; they and their fixed versions are in the versions table. | | an attacker can feed it Brotli-compressed input over the network | Three paths reach the decoder. (1) Inbound request bodies sent with Content-Encoding br: run `npm ls express body-parser` and look for Express 5 or body-parser 2.x, and for Koa with @koa/bodyparser. Per ELLIO, these inflate Brotli bodies before route handlers run. A bare node:http server does not decode request bodies by itself. (2) Outbound responses: built-in fetch(), undici, axios, got and SuperAgent decode Brotli responses, so any fetch to a URL an attacker influences counts (webhooks, link previews, crawlers). (3) Direct calls: `grep -rn "brotliDecompress\\|createBrotliDecompress" .` in your code and dependencies. Then check which services accept requests from untrusted networks, and whether your proxy or load balancer forwards Content-Encoding br to the application. Test your own staging service with a harmless Brotli body and read how it responds. Inventory features that fetch user-supplied URLs. | ## Affected and fixed versions | Product line | Affected versions | First fixed version | | --- | --- | --- | | Node.js 26.x | Before 26.12.1 | 26.12.1 | | Node.js 24.x | Before 24.22.1 | 24.22.1 | | Node.js 22.x | Before 22.23.4 | 22.23.4 | | Node.js 25.x, 23.x, 20.x and earlier (end of life) | All versions (Node.js: end-of-life versions are always affected) | No fix, move to a supported line | | Google Brotli 1.0.x | 1.0.3 up to but excluding 1.0.10 | 1.0.10 | | Google Brotli 1.1.x | 1.1.0 up to but excluding 1.1.1 | 1.1.1 | | Google Brotli 1.2.x | 1.2.0 up to but excluding 1.2.1 | 1.2.1 | | PyPI Brotli and brotlicffi | Brotli 1.0.3 up to but excluding 1.2.1; brotlicffi 1.0.3 up to but excluding 1.2.1.0 | 1.2.1 and 1.2.1.0 | | Siemens products with a fixed version in NVD | Cloud Connector Configurator before V1.24.3; Data Flow Monitoring before V3.0.18; Databus Configurator before V4.0.1; DataXess Configurator before V1.5.7; Edge OPC UA Server Configurator before V2.3.10; Flow Creator Configurator before V1.15.1; Gridscale X DSA V5.3, fixed in V5.4; Gridscale X PSS®E before V36.8.0; SIMATIC BRAUMAT and SIMATIC SISTAR V8.2 Update 1 up to before V8.2 Update 5; SIMATIC WinCC OA V3.20 before V3.20.17.1; SIMATIC WinCC OA V3.21 before V3.21.7; User Management Component (UMC) before V2.15.4.1 | The version NVD names as the upper bound (for example V1.24.3 for Cloud Connector Configurator); newer releases carry the fix | | Siemens products NVD lists as affected with no fixed version | Desigo CC V8 and V9, Industrial Edge Hub, Industrial Edge Management Cloud, Common Configurator (CC on IEM and CC on Edge), Insights Hub, Integrated PA Suite Tester, SENTRON Powermanager V8 and V9, SICAM PAS CC V6.02, SICAM SCC V7.01, HyperLynx, Innexis Architecture Native Acceleration 2026, PADS Pro Essentials, Student Edition and Professional, PAVE360 Automotive 2026, Xpedition Designer and Enterprise, Simcenter Executable Digital Twin and Studio, Process Preparation X, Teamcenter V2412, V2506, V2512 and V2606, Opcenter Connect Electronics and MOM, Opcenter Execution Core, Discrete, Foundation, Pharma and Process, Opcenter Intelligence, Intra Plant Logistics, Quality, RDnL and X, COMOS Walkinside, Project Server Web Administration Tool, SIMATIC Profiling Tool, WinCC Companion, WinCC Runtime Professional, WinCC Unified for Industrial Edge, WinCC Unified PC Runtime, WinCC Unified SIQENCE, WinCC V7/V8, and TIA Administrator | None named: ask Siemens support for patch status | Node.js releases are from the 21 Oct 2026 security release. Newer patch releases of the same lines also carry the fix. Siemens rows come from the NVD record for this CVE. ## If the fix cannot go in yet - Upgrade Node.js to the fixed release for your line in the versions table (or a newer patch release of the same line), then rebuild images and restart processes. Cost: a rebuild and redeploy. Node.js 20.x, 23.x and 25.x get no fix, so those need a move to a supported line. - Upgrade every other copy of Brotli: Google Brotli 1.0.10, 1.1.1 or 1.2.1, the distribution package where Node.js links the system library, Bun 1.4.4, PyPI Brotli 1.2.1. A Node.js upgrade fixes only the copy Node.js bundles. - The Node.js advisory lists no workaround. Stopgap from the researchers at Aikido and ELLIO: reject inbound requests with Content-Encoding br at the reverse proxy, load balancer or application middleware. Cost: clients that send Brotli-compressed request bodies get an error. This does not cover Brotli responses your own application fetches from other servers. - Vercel added firewall rules on 21 Oct 2026 that block the attack in inbound requests (reported by Aikido). Cost: none for Vercel-hosted apps; the rules cover inbound requests only. - CISA's KEV required action: apply mitigations per vendor instructions in line with BOD 26-04, or stop using the product if mitigations are unavailable. The KEV due date was 10 Nov 2026. ## What to look for - A failed attack against a patched build is rejected with ERR__ERROR_FORMAT_BLOCK_SWITCH (decoder error code -17) and a 400 response, per ELLIO. Search application logs for that string. - A failed attack against a vulnerable build often crashes the worker. Look for repeated Node.js process crashes or restarts, and for requests with Content-Encoding br just before them. On hosts with systemd-coredump, check /var/lib/systemd/coredump: ELLIO notes each crash can write a core file that holds credentials from other users' sessions, so limit who can read those files and how long you keep them. - Some other decoders (Python Brotli 1.2.0, brotlicffi 1.2.0.0, the PHP extension streaming API, the Ruby gem) return no bytes, never finish and never raise an error. Look for requests that hang with a CPU at 100 percent (ELLIO observed this in PHP). - Network indicators published by ELLIO: scanner user agent `Mozilla/5.0 (compatible; node2shell-check/1.0)` from AS44620 (sends 20 POST bodies with Content-Encoding br in two bursts after a GET to /); exploit traffic on 26 Oct 2026 from 94[.]154[.]43[.]7 carrying a 23,069,983-byte Brotli request body aimed at Node.js v24.22.0 on x86-64 (SHA-256 7fcbdaaecd4419e2930927844f3b772668204737a4e7602a2290c991c93c5c4b). - If a vulnerable, internet-reachable service saw such traffic or unexplained crashes, treat the host as possibly compromised and follow your incident process. CISA's KEV entry sets forensic triage to No for this CVE, so no BOD 26-04 triage steps are mandated for it. ## Timeline | Date | Event | Source | | --- | --- | --- | | 27 Feb 2026 | The upstream fix is committed to the public master branch of google/brotli (commit 0ed4268a, dated 27 Feb 2026) with no release and no advisory. | [source](https://ellio.tech/en/blog/node2shell-cve-2026-111497-rce-on-stock-nodejs-builds/) | | 14 Oct 2026 | Node.js sends a pre-alert of the 21 Oct security release, naming this flaw as high severity. | [source](http://www.openwall.com/lists/oss-security/2026/10/23/25) | | 21 Oct 2026 | Node.js releases 22.23.4, 24.22.1 and 26.12.1 and publishes the advisory. Brotli 1.2.1 is released the same day (1.0.10 and 1.1.1 also carry the fix). Vercel adds firewall rules for inbound requests. | [source](https://nodejs.org/en/blog/vulnerability/october-2026-security-releases) | | 22 Oct 2026 | NVD publishes the record. GitHub lists GHSA-9v8h-qqgm-4m35 (also the PyPI Brotli package). A public proof of concept appears, and Imperva's Red Team reports code execution on the official Node.js 24.22.0 arm64 image. | [source](https://www.aikido.dev/blog/node2shell-nodejs-brotli-rce-cve-2026-111497) | | 23 Oct 2026 | A post on oss-security points out that the flaw is in Google Brotli and reaches every consumer, and that Ubuntu rates it medium. | [source](http://www.openwall.com/lists/oss-security/2026/10/23/25) | | 25 Oct 2026 | ELLIO publishes its analysis: a scanner named node2shell-check is mapping hosts that decompress Brotli, and ELLIO reproduces code execution on x86-64 against four targets. | [source](https://ellio.tech/en/blog/node2shell-cve-2026-111497-rce-on-stock-nodejs-builds/) | | 26 Oct 2026 | ELLIO sensors record exploit traffic in the wild at 15:08 UTC, from 94[.]154[.]43[.]7, aimed at Node.js v24.22.0 on x86-64. | [source](https://ellio.tech/en/blog/node2shell-cve-2026-111497-rce-on-stock-nodejs-builds/) | | 27 Oct 2026 | CISA adds the CVE to the KEV catalog, due date 10 Nov 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-111497) | | 18 Nov 2026 | NVD last modified the record (status Analyzed, CVSS 3.1 score 9.8 from NVD). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-111497) | Exploited 5 days after disclosure: disclosed 21 Oct 2026, first exploitation reported 26 Oct 2026. ## Exploitation reported so far CISA added the CVE to KEV on 27 Oct 2026. The first public report of exploitation in the wild is from ELLIO: its sensors recorded a Brotli request body aimed at Node.js v24.22.0 on x86-64 on 26 Oct 2026 at 15:08 UTC, from 94[.]154[.]43[.]7. ELLIO says the delivery layer copies public proof-of-concept code and the final stage is new, and that its own test of this chain gave no shell in 50 attempts. ELLIO had seen the scanner node2shell-check probe for Brotli decoding the day before. Public exploits depend on the exact Node.js build, because they use fixed memory offsets for one binary and predictable socket numbers. Most failed attempts crash the worker, so even unsuccessful attacks cause denial of service. Aikido noted on 22 Oct that no exploitation in the wild had been reported yet. KEV does not know of ransomware use (Unknown). ## Why static reachability settles it Whether a vulnerable decoder runs on attacker-influenced data is a question about versions and code paths: which Node.js or library build you run, and which routes decompress request bodies or fetch outside URLs. Static inventory answers both. KEV tells you attackers are active, but only the inventory tells you which of your services can be hit. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-102764: BeyondTrust Remote Support](/ledger/cve-2026-102764/) - [CVE-2026-109619: PHP Group PHP](/ledger/cve-2026-109619/) - [CVE-2026-0322: Palo Alto Networks PAN-OS](/ledger/cve-2026-0322/) ## Sources - CVE-2026-111497: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-111497 (read 19 Nov 2026) - CVE-2026-111497: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-111497 (read 19 Nov 2026) - CVE-2026-111497: vendor advisory (Node.js security releases, 21 Oct 2026): https://nodejs.org/en/blog/vulnerability/october-2026-security-releases (read 19 Nov 2026) - CVE-2026-111497: Google Brotli 1.2.1 release notes (upstream fix release): https://github.com/google/brotli/releases/tag/v1.2.1 (read 19 Nov 2026) - CVE-2026-111497: GitHub advisory API record for the Brotli flaw (GHSA-9v8h-qqgm-4m35): https://api.github.com/advisories/GHSA-9v8h-qqgm-4m35 (read 19 Nov 2026) - CVE-2026-111497: ELLIO research on exploitation and affected builds: https://ellio.tech/en/blog/node2shell-cve-2026-111497-rce-on-stock-nodejs-builds/ (read 19 Nov 2026) - CVE-2026-111497: Aikido analysis and fix guidance: https://www.aikido.dev/blog/node2shell-nodejs-brotli-rce-cve-2026-111497 (read 19 Nov 2026) - CVE-2026-111497: oss-security mailing list post (23 Oct 2026): http://www.openwall.com/lists/oss-security/2026/10/23/25 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-111497/ --- # CVE-2026-102764: BeyondTrust Remote Support, is it exploitable where you run it Ledger entry, disclosed 15 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-102764 affects BeyondTrust Remote Support and was disclosed 15 Oct 2026 with a CVSS score of 9.3. Exposure needs these conditions: a self-hosted Remote Support appliance runs 26.2.3 or earlier without the BT26-05-RS patch; the appliance is reachable by the attacker over the network, with no credentials needed. In the CISA KEV catalog since 16 Oct 2026. ## What the flaw is The flaw is in internal messaging components of Remote Support. They did not adequately check where a message came from (CWE-290, improper origin validation). Under the right conditions, an attacker can use this to perform file operations on the appliance. Mandiant describes the result as unauthenticated remote file operations, including file upload, which can give host-level control and access to credentials the appliance stores. BeyondTrust has withheld technical details while self-hosted appliances are patched. The CNA rates it 9.3 on CVSS 4.0 (vector CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N); NVD adds a CVSS 3.1 base score of 9.8. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | a self-hosted Remote Support appliance runs 26.2.3 or earlier without the BT26-05-RS patch | BeyondTrust-hosted (SaaS) instances were patched by BeyondTrust and need no check; 26.3.1 and later include the fix. Sign in to the appliance administration interface at /appliance and read the installed version. Then look for BT26-05-RS among the applied updates, or confirm it was applied automatically if the appliance is subscribed to automatic updates. BeyondTrust publishes no build number for the patch, so presence in the applied list is the check. If the patch is not visible, open a BeyondTrust support case citing BT26-05. On a version older than 24.3.1 the patch does not apply: upgrade to a supported version first, then apply it. Self-hosted Privileged Remote Access shares the code: BeyondTrust found no exploit path there but released the patch as a precaution, so apply it too. | | the appliance is reachable by the attacker over the network, with no credentials needed | Check your firewall, reverse proxy or WAF rules for the appliance's published address and ports. Reachability from the internet is the highest-risk case. | ## Affected and fixed versions | Product line | Affected | Fixed | | --- | --- | --- | | Remote Support 24.3.1 to 26.2.3 | Yes | BT26-05-RS patch | | Remote Support older than 24.3.1 | Yes | Upgrade to a newer version, then apply the patch | | Remote Support 26.3.1 and greater | No | Fixed in the release | | Remote Support SaaS | Was affected | Patched by BeyondTrust (released 8 Oct 2026; all SaaS instances confirmed patched as of 9 Oct 2026) | | Privileged Remote Access | Shares code; no exploit path found | Patch released as a precaution | Source: BeyondTrust advisory BT26-05 (updated 4 Nov 2026) and the BeyondTrust blog. ## If the fix cannot go in yet - Expose only the parts of the appliance users need and restrict /login and /appliance to trusted networks. Cost: remote users and customers outside those networks lose access to the restricted parts. - Allow outbound traffic from the appliance only to required destinations. Cost: new integrations need an explicit allow rule. - Forward appliance syslog and session events to a SIEM in near real time. Cost: SIEM setup; the actor deletes local logs, so off-box copies are often the only record. ## What to look for - Files named by Mandiant: mangoapi.php (SAWFLY web shell) under the site service account's web root, /tmp/ingredi_ws (BARECLAW tunneler), /usr/bin/dbus-authd (unauthorized SUID-root binary), /home/bomgar/sbin/post_boot.d/zz-root-persist (persistence script), and a staging file named app_84304470.php. - Config changes: unexpected NOPASSWD entries in /ns/config/sudoers.template, and an unfamiliar ACL (wswsws) and backend (webskd3) in /ns/config/haproxy.conf. - Network: POST requests to a request path containing app_84304470, a WebSocket path /uidchanarge, a local listener on TCP 19800, and unknown .php request paths at the proxy or WAF. - Deleted or missing logs: /ns/log/auth.log, /ns/log/ssl.log, /ns/log/http-access.log, /var/log/messages, /home/bomgar/log/bt-secrets.log. Gaps in syslog forwarding that start and end within one session are a strong sign of tampering. - Syslog events: user_changed entries that raise jump_item_role or team roles from No Access to Administrator, and login success events on web/login from residential or commercial VPN addresses you cannot tie to an administrator or ticket. - Hashes and rules in the Mandiant post (IOC table and Detections section): SHA-256 dc14e215269a93fdbc4c1a80b984c23faf7af0911c6dbb85309796a6f2ae95fb for mangoapi.php (SAWFLY) and 1f19b9a27d79c11bda00a1e8faf74a7fdd683d38ccef1b60df2fd0307e3aa462 for ingredi_ws (BARECLAW); YARA rules G_Hunting_Webshell_SAWFLY_1 and G_Hunting_Backdoor_BARECLAW_1. The staging file app_84304470.php has no published hash. BeyondTrust offers a threat hunting package through its trust portal. The actor rotates infrastructure per victim, so hunt for behavior as well as indicators. No IP indicators were published. ## Timeline | Date | Event | Source | | --- | --- | --- | | 29 Aug 2026 | Suspected China-nexus actor exploits the flaw as a zero-day against one BeyondTrust-hosted Remote Support instance and one self-hosted appliance. | [source](https://www.beyondtrust.com/blog/entry/bt26-05) | | 28 Sep 2026 | BeyondTrust threat hunting finds the August intrusion on the SaaS instance and contains it within 27 minutes. | [source](https://www.beyondtrust.com/blog/entry/bt26-05) | | 30 Sep 2026 | Mandiant shares findings from the self-hosted customer. | [source](https://www.beyondtrust.com/blog/entry/bt26-05) | | 8 Oct 2026 | BeyondTrust releases the BT26-05 patch to SaaS and self-hosted customers and patches SaaS instances automatically. The advisory states all Remote Support SaaS customers were patched as of 9 Oct 2026. | [source](https://www.beyondtrust.com/blog/entry/bt26-05) | | 15 Oct 2026 | Coordinated disclosure: advisory BT26-05, CVE record, BeyondTrust blog and Mandiant analysis. NVD publishes the CVE. | [source](https://www.beyondtrust.com/trust-center/security-advisories/bt26-05) | | 16 Oct 2026 | CISA adds the CVE to KEV with a due date of 19 Oct 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-102764) | | 4 Nov 2026 | BeyondTrust updates advisory BT26-05. | [source](https://www.beyondtrust.com/trust-center/security-advisories/bt26-05) | Exploited 47 days before disclosure: disclosed 15 Oct 2026, first exploitation reported 29 Aug 2026. ## Exploitation reported so far BeyondTrust and Mandiant attribute the zero-day exploitation to a suspected China-nexus espionage actor that Mandiant calls ARMORY CASTLE (formerly UNC5221). BeyondTrust knows of two affected customers, one hosted and one self-hosted, and says it has seen no evidence of adversary activity since August. Mandiant expects broad exploitation of unpatched appliances now that the flaw is public, as the actor has done with past edge-device flaws. CISA marks ransomware use as Unknown. Response when an unpatched, internet-reachable appliance may be compromised (Mandiant): preserve evidence first by snapshotting virtual appliances with memory and collecting external syslog, proxy and firewall logs. Rebuild from a clean patched image rather than cleaning in place. Then rotate every credential the appliance stored or injected, in this order: /appliance admin, /login users, API client secrets, SAML and security provider keys, vault credentials, integration credentials. Review new MFA registrations on privileged accounts and open a Sev 1 case citing BT26-05. ## Why static reachability settles it Exposure depends on two facts you can read: the installed Remote Support version and whether the BT26-05 patch is applied, plus whether the appliance is reachable. Because exploitation preceded the patch, an exposed unpatched appliance also needs a compromise check using the detection list. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-0322: Palo Alto Networks PAN-OS](/ledger/cve-2026-0322/) - [CVE-2026-111497: Node.js Brotli decoder](/ledger/cve-2026-111497/) - [CVE-2026-97229: GitLab CE/EE, GitLab CI/CD regex RCE](/ledger/cve-2026-97229/) ## Sources - CVE-2026-102764: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-102764 (read 19 Nov 2026) - CVE-2026-102764: CISA KEV entry (KEV feed 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-102764 (read 19 Nov 2026) - CVE-2026-102764: vendor advisory BT26-05: https://www.beyondtrust.com/trust-center/security-advisories/bt26-05 (read 19 Nov 2026) - CVE-2026-102764: BeyondTrust blog on ARMORY CASTLE and BT26-05: https://www.beyondtrust.com/blog/entry/bt26-05 (read 19 Nov 2026) - CVE-2026-102764: Mandiant and GTIG analysis: https://cloud.google.com/blog/topics/threat-intelligence/armory-castle-exploits-beyondtrust (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-102764/ --- # CVE-2026-0322: Palo Alto Networks PAN-OS, is it exploitable where you run it Ledger entry, disclosed 14 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-0322 affects Palo Alto Networks PAN-OS and was disclosed 14 Oct 2026 with a CVSS score of 8.6. Exposure needs these conditions: PAN-OS runs below the fixed build for its release train; the attacker can reach the management interface and holds a valid administrator account. CISA added it to the Known Exploited Vulnerabilities catalog on 14 Oct 2026. ## What the flaw is CVE-2026-0322 is a code injection flaw (CWE-94) in the management interface of Palo Alto Networks PAN-OS. An attacker who can reach the management interface over the network and holds a valid account can run arbitrary code as root on the firewall. Root on a firewall gives full control of the device and the traffic it handles. It affects PAN-OS on PA-Series and VM-Series firewalls and on Panorama (virtual and M-Series). Cloud NGFW and Prisma Access are not affected. Palo Alto says no special configuration is needed to be affected, so every running instance of an affected build counts. Palo Alto rates it 8.6 on the CVSS 4.0 base score: network, low complexity, high privileges required, high impact on the device. NVD's own CVSS 3.1 score is 7.2 with the same high-privilege requirement. The vendor says risk is highest when the management interface is open to the internet, and that restricting it to trusted internal addresses minimizes the risk. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | PAN-OS runs below the fixed build for its release train | On the CLI run `show system info` and read the sw-version line. In the web interface, open Dashboard and read General Information > Software Version. Compare with the versions table. Check Panorama and every firewall it manages. `show system info` also prints the model and family lines. | | the attacker can reach the management interface and holds a valid administrator account | Open Device > Setup > Interfaces > Management Interface Settings and read Permitted IP Addresses. Also open Network > Network Profiles > Interface Mgmt and list every profile that allows HTTPS or SSH, then see which data-plane interfaces use it. From outside, test whether the management address answers from an untrusted network. A management address reachable from the internet is the highest-risk case. Open Device > Administrators and list accounts and their roles. Review Device > Authentication Profile and any RADIUS, LDAP or SAML server in use, and check who holds credentials, including shared or stale admin accounts. | ## Affected and fixed versions | Release train | Affected builds | Upgrade to | | --- | --- | --- | | PAN-OS 12.2 | 12.2.0 to 12.2.3 | 12.2.4 or later | | PAN-OS 12.1 | 12.1.8 up to, not including, 12.1.10-h2 | 12.1.10-h2 or later | | PAN-OS 12.1 | 12.1.5 up to, not including, 12.1.7-h8 | 12.1.7-h8, or 12.1.10-h2 or later | | PAN-OS 12.1 | 12.1.2 up to, not including, 12.1.4-h12 | 12.1.4-h12, or 12.1.10-h2 or later | | PAN-OS 11.2 | 11.2.14 to 11.2.15 | 11.2.16 or later | | PAN-OS 11.2 | 11.2.11 up to, not including, 11.2.13-h3 | 11.2.13-h3, or 11.2.16 or later | | PAN-OS 11.2 | 11.2.8 up to, not including, 11.2.10-h15 | 11.2.10-h15, or 11.2.16 or later | | PAN-OS 11.2 | 11.2.5 up to, not including, 11.2.7-h21 | 11.2.7-h21, or 11.2.16 or later | | PAN-OS 11.2 | 11.2.0 up to, not including, 11.2.4-h22 | 11.2.4-h22, or 11.2.16 or later | | PAN-OS 11.1 | 11.1.14 up to, not including, 11.1.16-h3 | 11.1.16-h3 or later | | PAN-OS 11.1 | 11.1.11 up to, not including, 11.1.13-h15 | 11.1.13-h15, or 11.1.16-h3 or later | | PAN-OS 11.1 | 11.1.8 up to, not including, 11.1.10-h35 | 11.1.10-h35, or 11.1.16-h3 or later | | PAN-OS 11.1 | 11.1.7 up to, not including, 11.1.7-h11 | 11.1.7-h11, or 11.1.16-h3 or later | | PAN-OS 11.1 | 11.1.5 up to, not including, 11.1.6-h39 | 11.1.6-h39, or 11.1.16-h3 or later | | PAN-OS 11.1 | 11.1.0 up to, not including, 11.1.4-h37 | 11.1.4-h37, or 11.1.16-h3 or later | | PAN-OS 10.2 | 10.2.17 up to, not including, 10.2.18-h12 | 10.2.18-h12 or later | | PAN-OS 10.2 | 10.2.14 up to, not including, 10.2.16-h11 | 10.2.16-h11, or 10.2.18-h12 or later | | PAN-OS 10.2 | 10.2.11 up to, not including, 10.2.13-h26 | 10.2.13-h26, or 10.2.18-h12 or later | | PAN-OS 10.2 | 10.2.8 up to, not including, 10.2.10-h41 | 10.2.10-h41, or 10.2.18-h12 or later | | PAN-OS 10.2 | 10.2.0 up to, not including, 10.2.7-h38 | 10.2.7-h38, or 10.2.18-h12 or later | | All older unsupported PAN-OS | All versions | Upgrade to a supported fixed version | Hotfix builds (-h) are listed per sub-train: any build below the fixed hotfix for your sub-train is affected, so pick the fixed build that matches your minor release. Siemens lists RUGGEDCOM APE1808 devices running a PAN-OS Virtual NGFW as affected by this CVE and points customers to Siemens support for patch information (Siemens SSA-104023, version 1.3, 10 Nov 2026). ## If the fix cannot go in yet - No workaround exists (Palo Alto). Upgrading to the fixed build for your train is the fix. - Restrict management interface access to trusted internal IP addresses, ideally one jump box that is the only system allowed to reach it. Cost: administrators must connect through that host or from those addresses, and any tool or script that manages the device from elsewhere stops working. Palo Alto rates the jump-box case 8.5 instead of 8.6 because the attack vector becomes adjacent, so it lowers the risk only a little and does not remove it. - Threat Prevention subscribers can enable Threat IDs 510046 and 510047 (Applications and Threats content version 9159-10287 or later). Palo Alto calls this limited coverage. It works only if you route management-port traffic through a data-plane interface (management profile on a data-plane interface), replace the certificate for inbound traffic management, decrypt inbound traffic to the management interface, and enable threat prevention on it. Cost: SSL decryption load and certificate work on the firewall, and attack variants may still pass. ## What to look for - Indicator addresses Palo Alto ties to limited exploit attempts: 66[.]42[.]72[.]14, 208[.]83[.]239[.]114, 149[.]248[.]47[.]234. In the web interface, open Monitor > Logs > System and filter on the source address in the event description: administrator logins to the management interface are recorded there with the client address. In Monitor > Logs > Configuration, the Host column holds the address an administrator connected from. In Monitor > Logs > Traffic, filter on the Source Address field, for example `( addr.src in 66.42.72.14 )`; this shows only sessions that crossed a data-plane interface, so it covers management traffic only where you route it through one. Palo Alto does not say which log records the attempts. - CISA's KEV entry marks forensic triage as required for this CVE. Its required action says to apply vendor mitigations in line with BOD 26-04 and CISA's forensics triage requirements. Palo Alto publishes no log signatures or file indicators beyond the three addresses. ## Timeline | Date | Event | Source | | --- | --- | --- | | 14 Oct 2026 | Palo Alto Networks publishes advisory CVE-2026-0322 (internal discovery, credited to Suraj Siddharudh) with fixed builds, exploit maturity ATTACKED and the note that very limited exploitation has been observed. NVD publishes the record and CISA adds the CVE to KEV the same day, with a due date of 17 Oct 2026. | [source](https://security.paloaltonetworks.com/CVE-2026-0322) | | 15 Oct 2026 | Palo Alto records an update to the advisory (the Updated field reads 2026-10-15). | [source](https://security.paloaltonetworks.com/CVE-2026-0322) | | 10 Nov 2026 | Siemens adds CVE-2026-0322 to advisory SSA-104023 for RUGGEDCOM APE1808 devices running PAN-OS Virtual NGFW (version 1.3). | [source](https://cert-portal.siemens.com/productcert/html/ssa-104023.html) | The earliest dated mark of exploitation is CISA's, on 14 Oct 2026, the day of disclosure; exploitation may have begun earlier. ## Exploitation reported so far CISA's SSVC data in the NVD record shows exploitation as active, automatable as no and technical impact as total. KEV lists ransomware campaign use as Unknown. ## Why attack-path analysis settles it Exposure needs two things at once: a network path to the management interface and a valid high-privilege account. The version row tells you if the code is present, but only the reachability and account rows tell you whether an attacker can use it. An attack-path view answers that, and the permitted-IP and administrator lists are where to read it. ## Related pages - [CVE-2026-97229: GitLab CE/EE, GitLab CI/CD regex RCE](/ledger/cve-2026-97229/) - [CVE-2026-102764: BeyondTrust Remote Support](/ledger/cve-2026-102764/) - [CVE-2026-21589: Atlassian Data Center products](/ledger/cve-2026-21589/) ## Sources - CVE-2026-0322: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-0322 (read 19 Nov 2026) - CVE-2026-0322: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-0322 (read 19 Nov 2026) - CVE-2026-0322: Palo Alto Networks security advisory (fixed builds, exploit status, Threat IDs, indicator addresses): https://security.paloaltonetworks.com/CVE-2026-0322 (read 19 Nov 2026) - CVE-2026-0322: Siemens ProductCERT advisory SSA-104023 (RUGGEDCOM APE1808): https://cert-portal.siemens.com/productcert/html/ssa-104023.html (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-0322/ --- # CVE-2026-97229: GitLab CE/EE, is it exploitable where you run it Ledger entry, disclosed 12 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-97229 affects GitLab CE/EE and was disclosed 12 Oct 2026 with a CVSS score of 10.0. Exposure needs the instance is self-managed GitLab CE or EE (not GitLab.com or GitLab Dedicated, which GitLab patched itself) and runs 19.2.0 to 19.2.7, 19.3.0 to 19.3.4 or 19.4.0 to 19.4.1. CISA added it to the Known Exploited Vulnerabilities catalog on 29 Oct 2026. ## What the flaw is GitLab lets users write CI/CD rules that match file paths with regular expressions. GitLab compiles and matches them with Ruby's Onigmo engine. In some cases the compiled form of a pattern grows past the size of a 32-bit integer and the size wraps around (CWE-190, integer overflow). The wrap lets the engine run past the end of its buffer, which an attacker can turn into code execution on the GitLab server. GitLab's detection guidance says the code runs as the git user. That user can read and write every repository on the instance and can read the secrets file, all CI/CD variables, runner tokens and access tokens. GitLab rates the flaw 10.0 (CVSS base score), with scope changed, and says an unauthenticated user can exploit it under certain conditions. GitLab has not published those conditions, and its detection guidance points at CI lint, GraphQL and pushed CI configuration, which in practice need an account, a token or a public project. Treat both as possible until GitLab says otherwise. This CVE is a bypass. GitLab 19.4.1 added input checks for the earlier CVE-2026-93577, and CVE-2026-97229 gets past them, so 19.4.1 is still affected. The Ruby project fixed the underlying engine bugs in its own releases of 22 and 23 Oct 2026 (4.0.8 on the 22nd; 4.0.9, 3.4.12 and 3.3.13 on the 23rd) and says they matter to any application that compiles regular expressions from attacker-controlled text. Johan Carlsson (joaxcar) reported the flaw through GitLab's bug bounty program. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the instance is self-managed GitLab CE or EE (not GitLab.com or GitLab Dedicated, which GitLab patched itself) and runs 19.2.0 to 19.2.7, 19.3.0 to 19.3.4 or 19.4.0 to 19.4.1 | Omnibus: head -1 /opt/gitlab/version-manifest.txt, or sudo gitlab-rake gitlab:env:info. Package: dpkg -l gitlab-ee gitlab-ce or rpm -q gitlab-ee gitlab-ce. Docker: docker exec head -1 /opt/gitlab/version-manifest.txt. Web UI: Help (/help) shows the version to signed-in users. Compare with 19.2.8, 19.3.5 and 19.4.2. Your own server, VM, container or Helm release is self-managed. | | an attacker can reach the web interface or API and can hold an account or token, or use a public project, to reach CI lint, GraphQL or pipeline processing; GitLab rates the flaw as needing no privileges but has not published the exact conditions, so treat any open path to those features as exposure | Find the address in /etc/gitlab/gitlab.rb (external_url). From a host outside your trusted network test with curl -sI , and review the firewall, load balancer and reverse proxy rules that publish the ports. GitLab's prevention advice is to limit who can create accounts and personal access tokens: in the Admin area open Settings > General > Sign-up restrictions and check whether sign-up is enabled and whether admin approval is required, open Visibility and access controls for the default project visibility, and review Admin > Users for accounts you do not recognize and Admin > Credentials for personal access tokens. | | a source install runs a system Ruby older than 4.0.8, 3.4.12 or 3.3.13 | Run ruby -v and compare with 4.0.8 (4.0.9 recommended), 3.4.12 or 3.3.13. Omnibus and Docker packages bundle a Ruby that the GitLab upgrade patches. | ## Affected and fixed versions | Product line | Affected | First fixed version | | --- | --- | --- | | GitLab 19.4 (CE and EE, self-managed) | 19.4.0 and 19.4.1 | 19.4.2 | | GitLab 19.3 (CE and EE, self-managed) | 19.3.0 to 19.3.4 | 19.3.5 | | GitLab 19.2 (CE and EE, self-managed) | 19.2.0 to 19.2.7 | 19.2.8 | | GitLab 19.1 and earlier | Not listed as affected by this CVE | Not applicable to this CVE | | GitLab.com and GitLab Dedicated | Patched by GitLab before the release | No action | | Ruby (source installs with a system Ruby) | Before 4.0.8, 3.4.12 or 3.3.13 for this flaw | 4.0.8 (released 22 Oct 2026), 3.4.12 and 3.3.13; Ruby recommends 4.0.9, 3.4.12 and 3.3.13 | GitLab 19.3.4 had an unrelated merge request page bug when the Ona integration was on, and 19.3.5 fixes it, so 19.3.5 is the one to install on that line. GitLab's own packages patch the Ruby they ship, so Omnibus and Docker users fix this by upgrading GitLab, not Ruby. ## If the fix cannot go in yet - Upgrade to GitLab 19.4.2, 19.3.5 or 19.2.8 or later, and prefer the newest patch of your minor version. 19.4.1 is not enough. GitLab's release notes list no workaround. Cost: a single-node upgrade restarts GitLab services, so plan a maintenance window and back up first. - Until the upgrade, limit who can create accounts and personal access tokens on the instance. This is GitLab's own prevention advice in its detection guidance. Cost: new users need an invitation or admin approval, and automation that creates tokens needs an administrator. It narrows who can reach the flaw but does not remove it. - Restrict network access to the GitLab web and API ports to trusted networks or a VPN. This is general hardening, not a GitLab instruction. Cost: contributors and runners outside those networks lose access. ## What to look for - GitLab released six detections for self-managed customers, published in its security-oss tldr repository. Workhorse log (Omnibus: /var/log/gitlab/gitlab-workhorse/current): a 502 with 'badgateway: failed to receive response: EOF' on POST /api/v4/projects/:id/ci/lint or /api/graphql, with a large read_bytes value and a correlation_id that is missing from api_json.log and graphql_json.log. GitLab calls this the highest-confidence signal. - application_json.log: WARN lines from Gitlab::Ci::Build::Rules::Rule::Clause::RegexpMatcher such as 'regexp rejected at compile time', 'rules:exists regexp match timed out' or 'regexp total time budget exceeded', with the pattern in the extra.regexp field. graphql_json.log: ciLint mutation requests whose logged variables carry oversized or unusual regular expression constructs. api_json.log: several 5xx responses on POST /api/v4/projects/:id/ci/lint from a raw scripted client such as python-urllib, python-requests, curl or Wget. - Host signs: [BUG] crash reports in puma_stderr.log (Omnibus: /var/log/gitlab/puma/puma_stderr.log) with frames from the org-ruby gem or Onigmo (onig_compile, compile_tree, match_at); 'A worker terminated, shutting down the cluster' in the Sidekiq log (/var/log/gitlab/sidekiq/current); API requests with a repository file path near the 255-byte limit that has a non-ASCII byte in its last segment. GitLab notes a successful exploit can be quiet, so a clean log does not prove safety. - If you find signs of code execution, GitLab says to treat the instance as compromised: rotate the secrets in gitlab-secrets.json, all CI/CD variables, runner tokens and access tokens. Hunt for files owned by git in /tmp, unexpected child processes of Puma or Sidekiq, and unexpected outbound connections. CISA's KEV entry marks forensic triage as Yes, so preserve the Workhorse, api_json, graphql_json and application_json logs first. ## Timeline | Date | Event | Source | | --- | --- | --- | | 12 Oct 2026 | GitLab releases critical patches 19.4.2, 19.3.5 and 19.2.8 for CE and EE. GitLab.com and Dedicated were already patched. | [source](https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-4-2-released/) | | 13 Oct 2026 | NVD publishes the record. CISA's SSVC entry on the record is stamped this day with exploitation active, automatable yes and technical impact total. This is the earliest dated public report of exploitation. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-97229) | | 14 Oct 2026 | A pull request adding a Metasploit exploit module for this CVE is opened. | [source](https://api.github.com/repos/rapid7/metasploit-framework/pulls/22017) | | 22 Oct 2026 | Ruby 4.0.8 is released. It carries the Ruby fix for the integer overflow in the size of a compiled pattern, the bug behind this CVE. | [source](https://www.ruby-lang.org/en/downloads/releases/) | | 23 Oct 2026 | The Ruby project releases 4.0.9, 3.4.12 and 3.3.13 and publishes its note on the engine fixes. Ruby's table lists the fix for this flaw as 4.0.8, 3.4.12 and 3.3.13. | [source](https://www.ruby-lang.org/en/news/2026/10/23/regexp-fixes-for-gitlab-cves/) | | 29 Oct 2026 | CISA adds the CVE to KEV with a due date of 1 Nov 2026 and forensic triage marked Yes. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-97229) | | 30 Oct 2026 | The Metasploit exploit module pull request is merged. | [source](https://api.github.com/repos/rapid7/metasploit-framework/pulls/22017) | | 17 Nov 2026 | The Ruby note is updated: Ruby treats the engine bugs as vulnerabilities and publishes its own advisories. | [source](https://www.ruby-lang.org/en/news/2026/10/23/regexp-fixes-for-gitlab-cves/) | The earliest dated mark of exploitation is CISA's, on 13 Oct 2026, 1 day after disclosure; exploitation may have begun earlier. ## Exploitation reported so far GitLab names no victims and no date of first use, and no vendor or researcher report describes an attack. A public exploit module titled as a GitLab RCE via CI Lint for this CVE is in the Metasploit framework, which lowers the effort for any attacker who can reach an unpatched instance. ## Why attack-path analysis settles it Exposure needs the build to be unfixed and then a route for an attacker to the CI lint, GraphQL or pipeline paths: network reach to the web and API ports, then an open sign-up, a token or a public project. The version read settles the first part. Mapping who can reach the instance and create accounts or tokens settles the second, which GitLab has not described. KEV and a public exploit module mean a reachable instance on an unfixed build should be patched and checked for compromise at once. ## Related pages - [CVE-2026-21589: Atlassian Data Center products](/ledger/cve-2026-21589/) - [CVE-2026-0322: Palo Alto Networks PAN-OS](/ledger/cve-2026-0322/) - [CVE-2026-88779: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-88779/) ## Sources - CVE-2026-97229: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-97229 (read 19 Nov 2026) - CVE-2026-97229: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-97229 (read 19 Nov 2026) - CVE-2026-97229: GitLab critical patch release notes 19.4.2, 19.3.5, 19.2.8 (vendor advisory): https://docs.gitlab.com/releases/patches/patch-release-gitlab-19-4-2-released/ (read 19 Nov 2026) - CVE-2026-97229: Ruby security note on regexp compilation fixes related to GitLab CVEs (updated 17 Nov 2026): https://www.ruby-lang.org/en/news/2026/10/23/regexp-fixes-for-gitlab-cves/ (read 19 Nov 2026) - CVE-2026-97229: GitLab security-oss detections for CVE-2026-97229 (GitLab detection guidance): https://gitlab.com/gitlab-security-oss/tldr/-/tree/main/Detections/platforms/GitLab/Vulnerability%20Detections (read 19 Nov 2026) - CVE-2026-97229: GitLab Workhorse worker crash detection (GitLab detection guidance): https://gitlab.com/gitlab-security-oss/tldr/-/raw/main/Detections/platforms/GitLab/Vulnerability%20Detections/ci_lint_workhorse_worker_crash.md (read 19 Nov 2026) - CVE-2026-97229: GitHub API record for Metasploit pull request 22017 (title and dates only): https://api.github.com/repos/rapid7/metasploit-framework/pulls/22017 (read 19 Nov 2026) - CVE-2026-97229: Ruby release list (4.0.8 dated 22 Oct 2026): https://www.ruby-lang.org/en/downloads/releases/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-97229/ --- # CVE-2026-21589: Atlassian Data Center products, is it exploitable where you run it Ledger entry, disclosed 5 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-21589 affects Atlassian Data Center products and was disclosed 5 Oct 2026 with a CVSS score of 9.3. Exposure needs these conditions: a self-managed Data Center, Crucible or Fisheye install runs below its fixed build; an attacker can reach the web application and no WAF or rewrite rule blocks parent-directory sequences. In the CISA KEV catalog since 15 Oct 2026. ## What the flaw is The listed Atlassian products serve web resources through a shared path-handling component. A crafted request can step outside the folder meant to be served and return other files from the web application root, such as configuration files under WEB-INF. Atlassian calls it arbitrary file access (CWE-552); the watchTowr researchers describe it as a pre-authentication arbitrary file read. The attacker needs no login. They must already know the exact name and path of the target file, and the flaw cannot list directories. How bad it gets depends on what sits in the web root. Atlassian says some configurations hold sensitive files there. watchTowr gave the example of the Crowd client file crowd.properties, which stores the application name and password that Jira, Confluence and others use to talk to Crowd. Atlassian rates it 9.3 (CVSS 4.0 base score; network, no privileges, high confidentiality impact on the product and high subsequent impact). The flaw reads files only; it does not write or run code by itself. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | a self-managed Data Center, Crucible or Fisheye install runs below its fixed build | Read the version from the product's administration System information page and compare it with the versions table. If you run the software yourself, it is in scope; Atlassian Cloud sites need no action, and Atlassian reported no evidence of exploitation there. | | an attacker can reach the web application and no WAF or rewrite rule blocks parent-directory sequences | From an untrusted network position, confirm the product's base URL answers through your reverse proxy, load balancer or firewall. Atlassian says instances reachable from the internet, including those with user authentication, should be restricted until you act. Then check your WAF or proxy for the vendor regex (Option 1 in Mitigation). On the host, check conf/server.xml for the RewriteValve line and WEB-INF/rewrite.config for the rule (Confluence, JSM, Jira, Bamboo, Crowd), or the first rule in app/WEB-INF/urlrewrite.xml (Bitbucket). | | sensitive files exist under the web application root | List what sits under WEB-INF and the other served folders (for example confluence/WEB-INF, atlassian-jira/WEB-INF, crowd-webapp/WEB-INF, app/WEB-INF) and look for credentials in config files, such as crowd.properties. Without such files the impact is lower, though the flaw is still present. | ## Affected and fixed versions | Product | Affected | First fixed (per supported release line) | | --- | --- | --- | | Bitbucket Data Center | All versions before the fix (introduced in 4.6.0). NVD ranges: 7.17.0 up to but excluding 9.4.26; 9.5.0 up to but excluding 10.2.8; 10.3.0 up to but excluding 10.5.1 | 9.4.26, 10.2.8, 10.5.1 | | Confluence Data Center | All versions before the fix (introduced in 5.10.0). NVD ranges: 5.10.0 up to but excluding 9.2.26; 9.3.0 up to but excluding 10.2.19 | 9.2.26, 10.2.19 | | Jira Service Management Data Center | All versions before the fix (introduced in 3.1.0). NVD ranges: 4.14.0 up to but excluding 5.12.40; 5.13.0 up to but excluding 10.3.26; 10.4.0 up to but excluding 11.3.12 | 5.12.40, 10.3.26, 11.3.12 | | Jira Software Data Center | All versions before the fix (introduced in 7.1.0). NVD ranges: 7.1.0 up to but excluding 9.12.40; 9.13.0 up to but excluding 10.3.26; 10.4.0 up to but excluding 11.3.12 | 9.12.40, 10.3.26, 11.3.12 | | Bamboo Data Center | All versions before the fix (introduced in 7.0.1). NVD ranges: 7.0.1 up to but excluding 10.2.24; 11.0.0 up to but excluding 12.1.12 | 10.2.24, 12.1.12 | | Crowd Data Center | All versions before the fix (introduced in 2.11.0). NVD ranges: 2.11.0 up to but excluding 6.3.7; 7.0.0 up to but excluding 7.0.3; 7.1.0 up to but excluding 7.1.7; 7.2.0 up to but excluding 7.2.4 | 6.3.7, 7.0.3, 7.1.7, 7.2.4 | | Crucible | All versions before the fix | 4.9.15 | | Fisheye | All versions before the fix | 4.9.15 | Atlassian's advisory table says 'All versions are affected' and tells you to patch to the fixed LTS version or later. If your release line is not listed, move to the nearest fixed line. The 'introduced' versions are from Atlassian's description in NVD. The NVD CPE ranges are narrower for Bitbucket (from 7.17.0) and Jira Service Management (from 4.14.0); older releases of those two are below the fixed builds and fall under the advisory's 'all versions'. Atlassian's security bug fix policy ships fixes as back-ported maintenance releases, not binary patches. Confluence: NVD ranges and Atlassian's fixed list stop at 10.2.19 (the LTS). Atlassian names no separate fix for Confluence 10.3 and later feature releases, states that all versions prior to the listed fixes are affected, and recommends patching to the fixed LTS version or the latest version, so on 10.3 or later move to the latest Confluence release. ## If the fix cannot go in yet - Upgrade to a fixed LTS version or later (versions table). Atlassian recommends removing the instance from the internet until it is patched. Cost: downtime for the upgrade, and removing public access stops external users and integrations. - Option 1, all products: block any URL that matches this pattern at the WAF or proxy: `(?is).*(?:/|\\|::|%(?:25)*(?:2f|5c)|(?::|%(?:25)*3a){2})(?:\.|%(?:25)*2e){2}(?:/|\\|::|%(?:25)*(?:2f|5c)|(?::|%(?:25)*3a){2}|;|%(?:25)*3b|$).*` The aim is to block two dots next to a slash, backslash or double colon, including URL-encoded and double-encoded forms. Test that the rule blocks those forms. Cost: any legitimate URL containing such a sequence will be denied. - Option 2, Confluence, JSM, Jira, Bamboo and Crowd: back up, then on each node shut it down, add `` inside the application's Context element in conf/server.xml (for Crowd, apache-tomcat/conf/Catalina/localhost/crowd.xml or apache-tomcat/conf/server.xml, whichever holds the Context), then add the vendor's rewrite.config rules to the product's WEB-INF directory (confluence/WEB-INF, atlassian-jira/WEB-INF, bamboo/WEB-INF or atlassian-bamboo/WEB-INF, crowd-webapp/WEB-INF) and restart. Each rule pairs a RewriteCond using the regex above against %{REQUEST_PATH} (and a second against %{REQUEST_URI} as a raw fallback) with `RewriteRule (?s)^.*$ /?denied [F,L]`. Cost: every node needs a restart, and denied URLs return a forbidden response. - Option 3, Bitbucket only: back up, then add a new rule as the first entry in app/WEB-INF/urlrewrite.xml on every node, mirror and mirror farm node, with the regex above in its from element, status 404, and target /mvc/error404, then restart Bitbucket. Cost: a restart on every node; matching requests receive a 404. ## What to look for - Atlassian cannot confirm whether an instance was affected and tells you to involve your security team to check all affected instances for signs of compromise. - Decoded pass: URL-decode each access-log request line, up to two passes, then look for two dots directly next to a slash, backslash or double colon. For example: `python3 -c "import sys,urllib.parse as u; [print(u.unquote(u.unquote(l)),end='') for l in sys.stdin]" < access.log | grep -E '(/|\\|::)\.\.(/|\\|::|;|$| )'`. A decoded hit looks like `GET ///../WEB-INF/ HTTP/1.1` with a 200 status and a response size above zero. - Raw pass: search the undecoded lines with the vendor regex, which also matches single and double encoded forms: `grep -Pi '(?:/|\\|::|%(?:25)*(?:2f|5c)|(?::|%(?:25)*3a){2})(?:\.|%(?:25)*2e){2}(?:/|\\|::|%(?:25)*(?:2f|5c)|(?::|%(?:25)*3a){2}|;|%(?:25)*3b|$)' access.log`. - For hits that returned content from WEB-INF, rotate any credentials stored in files under the web root, for example the Crowd application password in crowd.properties. - Atlassian and CISA have published no indicator addresses or file hashes. ## Timeline | Date | Event | Source | | --- | --- | --- | | 5 Oct 2026 | Atlassian publishes the advisory and fixed versions; NVD publishes the record | [source](https://confluence.atlassian.com/security/cve-2026-21589-arbitrary-file-access-vulnerability-impacts-multiple-products-1870495748.html) | | 6 Oct 2026 | watchTowr Labs publishes technical research on the flaw | [source](https://labs.watchtowr.com/you-wont-hear-about-these-even-in-myths-atlassian-jira-confluence-and-more-pre-auth-arbitrary-file-read-cve-2026-21589/) | | 15 Oct 2026 | CISA adds the CVE to KEV (forensic triage required) with a due date of 18 Oct 2026 | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-21589) | | 29 Oct 2026 | NVD record last modified (status Analyzed) | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-21589) | The earliest dated mark of exploitation is CISA's, on 15 Oct 2026, 10 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far Neither CISA nor Atlassian gives an exploitation date for Data Center, and KEV lists ransomware use as unknown. For Cloud, Atlassian says its investigation found no evidence of exploitation. ## Why static reachability settles it Whether you are exposed comes down to checkable facts: the product version against the fixed build, whether the web application is reachable, whether a rewrite or WAF rule is in place, and whether sensitive files sit under the web root. Each can be read from the host or the network without running an attack. Because it is known exploited, an affected reachable instance should be patched first and then reviewed for exposed credentials. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-88779: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-88779/) - [CVE-2026-97229: GitLab CE/EE, GitLab CI/CD regex RCE](/ledger/cve-2026-97229/) - [CVE-2026-104286: Fortinet FortiMail, FortiMail path traversal](/ledger/cve-2026-104286/) ## Sources - CVE-2026-21589: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-21589 (read 19 Nov 2026) - CVE-2026-21589: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-21589 (read 19 Nov 2026) - CVE-2026-21589: Atlassian security advisory (vendor advisory, all eight products): https://confluence.atlassian.com/security/cve-2026-21589-arbitrary-file-access-vulnerability-impacts-multiple-products-1870495748.html (read 19 Nov 2026) - CVE-2026-21589: watchTowr Labs research, 6 Oct 2026: https://labs.watchtowr.com/you-wont-hear-about-these-even-in-myths-atlassian-jira-confluence-and-more-pre-auth-arbitrary-file-read-cve-2026-21589/ (read 19 Nov 2026) - CVE-2026-21589: GitHub advisory database entry GHSA-gr66-x5pq-8g2g: https://api.github.com/advisories?cve_id=CVE-2026-21589 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-21589/ --- # CVE-2026-88779: Citrix NetScaler ADC and Gateway, is it exploitable where you run it Ledger entry, disclosed 3 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-88779 affects Citrix NetScaler ADC and Gateway and was disclosed 3 Oct 2026 with a CVSS score of 8.7. Exposure needs these conditions: the NetScaler build is before the fixed build for its train; the appliance is configured as a SAML service provider (SP) or identity provider (IdP). In the CISA KEV catalog since 4 Oct 2026. ## What the flaw is CVE-2026-88779 is a memory overflow (CWE-119, improper restriction of operations within the bounds of a memory buffer) in NetScaler ADC and NetScaler Gateway. Citrix says it causes denial of service. The CVSS 4.0 score is 8.7: network attack, low complexity, no privileges, no user interaction, with high impact on availability only and none on confidentiality or integrity. The flaw is reachable only when the appliance is set up as a SAML SP or a SAML IdP. An appliance with neither configured does not meet the vendor's precondition. The bulletin covers customer-managed appliances only; Citrix-managed cloud services and Citrix-managed Adaptive Authentication are updated by Cloud Software Group. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | The NetScaler build is before the fixed build for its train | On the appliance CLI run `show ns version` and compare with the versions table. Check every node of an HA pair or cluster. | | The appliance is configured as a SAML service provider (SP) or identity provider (IdP) | Citrix says to inspect the configuration for entries matching `add authentication samlAction` (SAML SP) or `add authentication samlIdPProfile` (SAML IdP). Run `show ns runningConfig \| grep "add authentication samlAction"` and `show ns runningConfig \| grep "add authentication samlIdPProfile"`. A hit on either command means the condition holds. | | The SAML endpoint is reachable by the attacker, who Citrix and CISA describe as unauthenticated and remote | Find the virtual servers that use the SAML action or IdP profile, then check which of their addresses are published to the internet or untrusted networks. On the appliance CLI, list the SAML objects with `show ns runningConfig \| grep -E "samlAction\|samlIdPProfile\|samlPolicy\|samlIdPPolicy"`, then find where they are attached with `show ns runningConfig \| grep -E "bind (authentication\|vpn\|aaa) vserver"` and read each virtual server's address with `show vserver ` (or `show authentication vserver `). A SAML virtual server whose address is reachable from the internet, or through a published URL, meets the condition. These are general NetScaler configuration commands; Citrix names only the two `add authentication` entries as the test. | ## Affected and fixed versions | Product line | Affected builds | First fixed build | | --- | --- | --- | | NetScaler ADC and Gateway 14.1 | Before 14.1-73.41 | 14.1-73.41 | | NetScaler ADC and Gateway 13.1 | Before 13.1-64.28 | 13.1-64.28 | | NetScaler ADC 14.1-FIPS | Before 14.1-73.41 FIPS | 14.1-73.41 FIPS | | NetScaler ADC 13.1-FIPS and 13.1-NDcPP | Before 13.1-37.282 | 13.1-37.282 | Each fixed build and all later releases of the same train are fixed. The 13.1-FIPS and 13.1-NDcPP builds are a separate build stream from the standard 13.1 builds, which is why bulletin CTX697174 gives 13.1-37.282 for them and 13.1-64.28 for standard 13.1; compare a FIPS appliance only with its own row. Secure Private Access Hybrid deployments need their NetScaler instances moved to these builds too. ## If the fix cannot go in yet - No workaround is listed in bulletin CTX697174; the only change Citrix offers is an upgrade to the first fixed build for your train. The bulletin does not state downtime, so plan for the restart an ADC upgrade normally needs. - HA pair upgrade order (NetScaler documentation, page last published 12 Nov 2026): upgrade the secondary node first, run `force failover` so it becomes primary, then upgrade the original primary and fail over again if you want the original roles back. Both nodes must end on the same release. While the two nodes run different builds with different internal HA versions, HA configuration sync, command propagation, state sync and connection mirroring are disabled, so make changes and run the pair in that state only for the length of the upgrade. - Exposure-reducing step from the vendor's precondition: the flaw needs a SAML SP or IdP configuration. If neither is configured the appliance does not meet the precondition, but removing a SAML configuration stops SAML sign-in through that appliance, so it is a choice for you to weigh, not a vendor-recommended workaround. - CISA's KEV entry lists the required action: apply mitigations in accordance with vendor instructions, or discontinue use of the product if mitigations are unavailable. ## What to look for - Citrix and CISA publish no log strings or indicator addresses for this flaw, and CISA marks forensic triage as not required. The impact is denial of service, so the trace an appliance leaves is a restart or crash. On the appliance shell, look for new crash dump directories under /var/core, restart and error lines around the event time in /var/log/ns.log, and a recent start time in the `show ns stats` output (System up since). These are general NetScaler places to look, not published indicators for this CVE. ## Timeline | Date | Event | Source | | --- | --- | --- | | 3 Oct 2026 | Citrix publishes bulletin CTX697174 (initial publication, 3 Oct 2026 PST) with the fixed builds and credits JPMorgan Chase, Bishop Fox and watchTowr. | [source](https://support.citrix.com/external/article/CTX697174/netscaler-adc-and-netscaler-gateway-secu.html) | | 4 Oct 2026 | NVD publishes the record at 04:16 UTC (21:16 Pacific daylight time on 3 Oct 2026). CISA adds the CVE to KEV with a due date of 7 Oct 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-88779) | The earliest dated mark of exploitation is CISA's, on 4 Oct 2026, 1 day after disclosure; exploitation may have begun earlier. ## Exploitation reported so far No report names the attacker or the first attack. Citrix's 3 Oct 2026 date is Pacific time and CISA publishes a date with no time, so the one-day gap is a calendar-date difference, not a measured interval. ## Why static reachability settles it Exposure depends on the build and on one configuration fact: whether the appliance acts as a SAML SP or IdP. Both are readable from the appliance with `show ns version` and a grep of the running configuration. KEV shows attacks happened, but only the configuration read tells you whether your appliance can be hit. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-104286: Fortinet FortiMail, FortiMail path traversal](/ledger/cve-2026-104286/) - [CVE-2026-21589: Atlassian Data Center products](/ledger/cve-2026-21589/) - [CVE-2026-94545: Vercel Satori and Next.js](/ledger/cve-2026-94545/) ## Sources - CVE-2026-88779: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-88779 (read 19 Nov 2026) - CVE-2026-88779: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-88779 (read 19 Nov 2026) - CVE-2026-88779: Citrix security bulletin CTX697174 (vendor advisory): https://support.citrix.com/external/article/CTX697174/netscaler-adc-and-netscaler-gateway-secu.html (read 19 Nov 2026) - CVE-2026-88779: NetScaler documentation, upgrade a high availability pair: https://docs.netscaler.com/en-us/citrix-adc/current-release/upgrade-downgrade-citrix-adc-appliance/upgrade-downgrade-ha-pair (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-88779/ --- # CVE-2026-104286: Fortinet FortiMail, is it exploitable where you run it Ledger entry, disclosed 1 Oct 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-104286 affects Fortinet FortiMail and was disclosed 1 Oct 2026 with a CVSS score of 9.8. Exposure needs these conditions: self-managed FortiMail (not FortiMail Cloud) runs an affected release; the IBE service is enabled (inferred from Fortinet's workaround and log lines; the advisory does not state it as a condition) and the web interface is reachable by the attacker. ## What the flaw is FortiMail has a path traversal flaw (CWE-22) combined with improper handling of a NULL byte (CWE-158) in its web interface. By sending crafted HTTP or HTTPS requests, an unauthenticated attacker can write arbitrary files on the underlying system. Fortinet lists the impact as execution of unauthorized code or commands, and rates it 9.8 (CVSS base score; temporal 8.8). Fortinet's workaround and its log lines tie the flaw to the IBE service. Writing files on a mail gateway can lead to code running as root. The indicators Fortinet publishes show what attackers did after that: a cron job run by root, an admin session logged out for no clear reason, and a new mail archive account that sends archived mail to an outside server by remote copy. That last step would give an attacker a copy of mail passing through the appliance. Fortinet's own Product Security team found the flaw. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | self-managed FortiMail (not FortiMail Cloud) runs an affected release | In the CLI run get system status and read the Version line. In the GUI open Dashboard > Status and read System Information. Compare with the versions table. On a FortiMail Cloud instance the same command should report 7.6.7 GA or 8.0.2 GA. | | the IBE service is enabled (inferred from Fortinet's workaround and log lines; the advisory does not state it as a condition) and the web interface is reachable by the attacker | IBE: in the GUI open Encryption > IBE and read IBE Service; in the CLI run show system encryption ibe (or show full-configuration system encryption ibe to include defaults) and look for set status enable. Reachability: run show system interface and read set allowaccess on each interface for http and https, then from outside the trusted network test with nc -vz 443 (and 80). Review firewall and NAT rules that publish the appliance, including webmail and the IBE portal. | ## Affected and fixed versions | FortiMail train | Affected | First fixed release | | --- | --- | --- | | 8.0 | 8.0.0 through 8.0.1 | 8.0.2 | | 7.6 | 7.6.0 through 7.6.6 | 7.6.7 | | 7.4 | 7.4.0 through 7.4.8 | 7.4.9 | | 7.2 | 7.2.0 through 7.2.9 | 7.2.10 (listed as upcoming in the advisory revision of 7 Oct 2026) | | 7.0 | All versions | No fixed 7.0 release; migrate to a fixed release | | FortiMail Cloud | All firmware before 5 Oct 2026 | Fixed by Fortinet on 5 Oct 2026 by updating every system to 7.6.7 GA or 8.0.2 GA; no customer action | NVD's product list shows 7.0.0 through 7.0.9, while Fortinet's advisory says all 7.0 versions are affected. Use Fortinet's statement. ## If the fix cannot go in yet - Turn off the IBE service. In the GUI: Encryption > IBE > IBE Service set to off. In the CLI: config system encryption ibe, then set status disable, then end. Cost: IBE secure mail stops working. Recipients can no longer open or reply to IBE-encrypted messages through the FortiMail secure portal, and users who depend on IBE policies lose that delivery path. - Or block internet access to the FortiMail webmail interface and allow it only from trusted private networks. Cost: external users, including IBE recipients, lose access to webmail and the secure portal. - Or, if a web application firewall sits in front of FortiMail, block POST requests to /ibe whose content contains a parent-directory sequence (../). Cost: the rule has to be written and tested on your WAF. - Fortinet urges customers who cannot upgrade to apply one of these workarounds because exploitation is reported. ## What to look for - Look for signs that the appliance is already compromised before and after patching. Indicator addresses published by Fortinet: 79[.]141[.]169[.]187 and 45[.]129[.]0[.]192. Search firewall and proxy logs for connections to them, and in the GUI check the Archive account settings for entries you did not create (archive234 in Fortinet's example). - System event log (Log > Event > System) lines Fortinet lists: a cron entry run as root, shown as type=event subtype=system pri=debug user=system ui=cron msg="(root) CMD (/bin/sh -c 'O=/migadmin ..."; an admin logout with an unknown reason, type=kevent subtype=admin pri=information user=admin ui=(null) action=logout status=success reason=unknown msg="User admin logged out from (null)."; and a configuration entry that adds an archive account with a remote destination, type=kevent subtype=config ... msg="Added 'archive234' to 'archive account'" with rotation-size[50] rotation-time[1] rotation-hour[14] destination[remote] remote-ip[79[.]141[.]169[.]187] remote-username[archive234] remote-directory[/uploads] (user: admin, from: cli). - Encryption log (Log > Encryption) lines Fortinet lists: FortiMail::IBE::DecrypterMediaIn constructor entries that read "Caught BufferException(2), BufferImpl.cpp:973, 'Invalid Base64 Encoding at pos 0. Character=0x2a'", followed by "Internal user *@domain.tld failed to log in.". - CISA's KEV entry marks forensic triage as Yes for this CVE: preserve the appliance logs and a configuration backup before you change the appliance. ## Timeline | Date | Event | Source | | --- | --- | --- | | 1 Oct 2026 | Fortinet publishes advisory FG-IR-26-175 with the workaround and states the flaw is exploited in the wild. NVD publishes the record and CISA adds the CVE to KEV the same day, with a due date of 4 Oct 2026. | [source](https://fortiguard.fortinet.com/psirt/FG-IR-26-175) | | 2 Oct 2026 | CISA's SSVC entry on the NVD record is stamped this day: exploitation active, automatable yes, technical impact total. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-104286) | | 5 Oct 2026 | Fortinet revises the advisory (solution update) and updates FortiMail Cloud to 7.6.7 GA or 8.0.2 GA. | [source](https://fortiguard.fortinet.com/psirt/cvrf/FG-IR-26-175) | | 7 Oct 2026 | Fortinet revises the advisory again (FortiMail Cloud fix clarification). It lists FortiMail 7.2.10 as upcoming. | [source](https://fortiguard.fortinet.com/psirt/cvrf/FG-IR-26-175) | The earliest dated mark of exploitation is CISA's, on 1 Oct 2026, the day of disclosure; exploitation may have begun earlier. ## Exploitation reported so far The advisory names no reporter of the attacks and no date for the first one, and Fortinet has not said how many appliances were hit. KEV lists ransomware campaign use as Unknown. ## Why static reachability settles it Exposure needs an affected FortiMail build, the IBE service on, and a web interface the attacker can reach. A version read, an IBE setting read and a reachability test answer all three on each appliance. Fortinet and CISA confirm active use, so any appliance that passes all three checks should be treated as possibly compromised and checked against the log indicators. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-94545: Vercel Satori and Next.js](/ledger/cve-2026-94545/) - [CVE-2026-88779: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-88779/) - [CVE-2026-94127: F5 BIG-IP APM, BIG-IP APM OAuth server RCE](/ledger/cve-2026-94127/) ## Sources - CVE-2026-104286: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-104286 (read 19 Nov 2026) - CVE-2026-104286: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-104286 (read 19 Nov 2026) - CVE-2026-104286: Fortinet PSIRT advisory FG-IR-26-175 (vendor advisory): https://fortiguard.fortinet.com/psirt/FG-IR-26-175 (read 19 Nov 2026) - CVE-2026-104286: Fortinet PSIRT advisory FG-IR-26-175 in CVRF form (revision history): https://fortiguard.fortinet.com/psirt/cvrf/FG-IR-26-175 (read 19 Nov 2026) - CVE-2026-104286: FortiMail CLI reference, system encryption ibe: https://docs.fortinet.com/document/fortimail/7.6.0/cli-reference/6402/system-encryption-ibe (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-104286/ --- # CVE-2026-94545: Vercel Satori and Next.js, is it exploitable where you run it Ledger entry, disclosed 30 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-94545 affects Vercel Satori and Next.js and was disclosed 30 Sep 2026 with a CVSS score of 5.3. Exposure needs these conditions: the app uses Satori or the Node.js next/og on an affected version; attacker-controlled values reach the generated SVG. It was not in the CISA Known Exploited Vulnerabilities catalog on the read date. ## What the flaw is Satori turns HTML and CSS into SVG. It does not properly escape some values before it writes them into the SVG output, so a crafted value can be read as SVG markup instead of plain text. In the Satori advisory, the impact depends on how the generated SVG is consumed. Next.js ships Satori inside the Node.js implementation of ImageResponse in next/og. The Next.js advisory calls this an upstream vulnerability and rates it as remote code execution. It applies when an application passes attacker-controlled values into SVG content, attributes or styles during image generation. A typical case is a route that reads a query parameter and places it in the SVG. The Edge implementation of ImageResponse is not affected, and neither are applications that never pass attacker-controlled values into the SVG. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the app uses Satori or the Node.js next/og on an affected version | Run `npm ls satori next` in the project (or `pnpm why satori`), and `node -p "require('next/package.json').version"` in the deployed app directory; compare with the versions table. Then find the use: `grep -rn "next/og" app pages src` and `grep -rn "satori" package.json`. Open Graph image routes (opengraph-image files) and image API routes are the usual places. | | attacker-controlled values reach the generated SVG | Condition in full: attacker-controlled values reach the SVG content, attributes or styles during image generation; for Satori used alone, the SVG must also be read by something that interprets SVG markup. In each ImageResponse route, trace every value placed inside the JSX tree: request URL parameters, headers, cookies and database fields that users can write. Vulnerable shape (the vendor's own example): a GET handler reads a query parameter with `new URL(request.url).searchParams.get('value')` and renders it as `{value}`. Safe shapes: a route that renders only fixed text, or a value your own code picks from a fixed list, passes no attacker-controlled value into the SVG. For Satori used alone, find where the SVG string goes after Satori returns it: the Satori advisory states the impact depends on how the SVG is consumed, and output rasterized to PNG on your side differs from SVG served to browsers or passed to another service. | | the ImageResponse route runs on the Node.js runtime, not Edge | Run `grep -rn "export const runtime" app pages src`. A route with `export const runtime = 'edge'` uses the Edge implementation, which the vendor says is not affected, while a route with no runtime export, or with `export const runtime = 'nodejs'`, is on Node.js and is in scope. Moving a route to Edge means adding `export const runtime = 'edge'` to that route file; Edge runtime limits then apply to it. | ## Affected and fixed versions | Package | Affected versions | First fixed version | | --- | --- | --- | | satori | 0.0.27 up to but excluding 0.33.5 | 0.33.5 | | next (Node.js ImageResponse from next/og only) | 16.2.0 up to but excluding 16.3.6 | 16.3.6 | The Next.js advisory lists no Next.js version before 16.2.0 as affected. Next.js 16.3.6 and satori 0.33.5 were both released on 22 Sep 2026, eight days before the Next.js advisory was published on 30 Sep 2026; the Satori advisory, which carries the CVE ID, went public on 27 Oct 2026 (GitHub advisory API, read 19 Nov 2026). The fix is to upgrade to next 16.3.6 and satori 0.33.5 (cost: a rebuild, a redeploy and a check that generated images still look right). The two ratings come from two advisories: the Next.js advisory GHSA-vcvr-r3jv-pc5j rates the Next.js impact (remote code execution) 9.5, because attacker-controlled values in the SVG there lead to code execution; the Satori advisory GHSA-wx4j-mvgx-mqwp rates the Satori library flaw alone 5.3, because its impact depends on how the SVG is consumed. The CVE record and NVD carry the 5.3. GitHub's advisory API shows no CVE ID on the Next.js advisory. ## If the fix cannot go in yet - Next.js workaround from the vendor: if you cannot upgrade, do not pass attacker-controlled values into SVG content, attributes or styles rendered by the Node.js ImageResponse. Cost: dynamic text taken from users (titles, names, query parameters) must be replaced with fixed values or values you control, so personalized images stop working. - Satori workaround from the vendor: no complete workaround exists besides upgrading. Applications that cannot upgrade should not render attacker-controlled content with Satori. Cost: the same loss of user-supplied content in generated images. - Switching a route to the Edge ImageResponse implementation is outside the workaround the vendor lists, but the vendor states the Edge implementation is not affected. Cost: Edge runtime limits apply to that route. ## What to look for - Run the version and code checks from the checks table on every service. No indicators, log strings or file paths are published by the vendor or NVD for this CVE, and the vendor does not name a log source to check. - Suggested hunt (not published by the vendor): in web server or CDN logs for your Open Graph and image routes, look for query-parameter values that contain markup characters, such as `<`, `>`, `%3C`, `%3E`, or quotes, which a normal title or name does not hold. ## Timeline | Date | Event | Source | | --- | --- | --- | | 22 Sep 2026 | Satori 0.33.5 is published to npm with the fix. | [source](https://registry.npmjs.org/satori) | | 22 Sep 2026 | Next.js 16.3.6 is released with the fix for the next/og ImageResponse, eight days before the Next.js advisory is public. | [source](https://github.com/vercel/next.js/releases/tag/v16.3.6) | | 30 Sep 2026 | GitHub publishes the Next.js advisory GHSA-vcvr-r3jv-pc5j at 14:48 UTC (critical, CVSS 4.0 score 9.5). GitHub's advisory API shows no CVE ID on it. | [source](https://api.github.com/advisories/GHSA-vcvr-r3jv-pc5j) | | 30 Sep 2026 | GitHub, as CNA, publishes the CVE record and NVD lists it at 15:22 UTC with the CVSS 4.0 score 5.3. The NVD references link to both advisories and the fix commits; the Satori advisory page was not yet public. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-94545) | | 27 Oct 2026 | GitHub makes the Satori advisory GHSA-wx4j-mvgx-mqwp public at 14:46 UTC (severity medium, CVSS 4.0 score 5.3, patched in 0.33.5). It is the advisory that carries this CVE ID. | [source](https://api.github.com/advisories/GHSA-wx4j-mvgx-mqwp) | No public report dates the start of exploitation. ## Why static reachability settles it Exposure needs an affected version on the Node.js ImageResponse and a data flow from user input into the SVG tree. Both are facts about the code, so tracing the data flow in the source settles it. A severity score cannot tell you this, since the two vendors rate the same flaw 9.5 and 5.3. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-94127: F5 BIG-IP APM, BIG-IP APM OAuth server RCE](/ledger/cve-2026-94127/) - [CVE-2026-104286: Fortinet FortiMail, FortiMail path traversal](/ledger/cve-2026-104286/) - [CVE-2026-76460: Cisco Identity Services Engine](/ledger/cve-2026-76460/) ## Sources - CVE-2026-94545: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-94545 (read 19 Nov 2026) - CVE-2026-94545: vendor advisory (Next.js GHSA-vcvr-r3jv-pc5j, via the GitHub advisory API): https://github.com/vercel/next.js/security/advisories/GHSA-vcvr-r3jv-pc5j (read 19 Nov 2026) - CVE-2026-94545: GitHub advisory API record for the Next.js advisory: https://api.github.com/advisories/GHSA-vcvr-r3jv-pc5j (read 19 Nov 2026) - CVE-2026-94545: vendor advisory (Satori GHSA-wx4j-mvgx-mqwp, via the GitHub advisory API): https://github.com/vercel/satori/security/advisories/GHSA-wx4j-mvgx-mqwp (read 19 Nov 2026) - CVE-2026-94545: GitHub advisory API record for the Satori advisory: https://api.github.com/advisories/GHSA-wx4j-mvgx-mqwp (read 19 Nov 2026) - CVE-2026-94545: Next.js 16.3.6 release notes: https://github.com/vercel/next.js/releases/tag/v16.3.6 (read 19 Nov 2026) - CVE-2026-94545: npm registry record for satori (publish time of 0.33.5): https://registry.npmjs.org/satori (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-94545/ --- # CVE-2026-94127: F5 BIG-IP APM, is it exploitable where you run it Ledger entry, disclosed 22 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-94127 affects F5 BIG-IP APM and was disclosed 22 Sep 2026 with a CVSS score of 9.3. Exposure needs these conditions: APM runs 21.1.0, 17.5.x or 17.1.x without the matching hotfix; APM acts as an OAuth Authorization Server on a virtual server the attacker can reach. CISA added it to the Known Exploited Vulnerabilities catalog on 22 Sep 2026. ## What the flaw is CVE-2026-94127 is a heap-based buffer overflow (CWE-122) in BIG-IP Access Policy Manager (APM). When an access policy and an OAuth profile are configured on a virtual server, specific malicious traffic can cause remote code execution. No authentication is needed. F5 rates it 9.3 on CVSS 4.0 (the rating used throughout this page) and 9.8 on CVSS 3.1. The flaw is present only when APM is set up as an OAuth Authorization Server. Deployments that use APM strictly as an OAuth client or resource server, without authorization server profiles, are not affected. The problem sits in the data plane, so management interfaces are not the route in. F5 states that systems in Appliance mode are also vulnerable. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | APM runs 21.1.0, 17.5.x or 17.1.x without the matching hotfix | The hotfixes are Hotfix-BIGIP-21.1.0.2.0.30.22-ENG (21.1), Hotfix-BIGIP-17.5.1.9.0.160.12-ENG (17.5) and Hotfix-BIGIP-17.1.3.5.0.41.14-ENG (17.1). On the BIG-IP run `tmsh show sys version` for the base version and build, and `tmsh show sys software status` to see installed images and hotfixes on each volume. In the GUI use System > Software Management. Compare with the versions table. Check each device in a failover pair. | | APM acts as an OAuth Authorization Server on a virtual server the attacker can reach | An APM access policy and an OAuth profile must both be attached to the virtual server; a client-only or resource-server-only setup is not affected. List the OAuth profiles with `tmsh list apm oauth profile`; an empty result means no OAuth profile exists and F5's condition does not hold. Then run `tmsh list ltm virtual all-properties` and look for a virtual server whose profiles include both an access profile and one of those OAuth profiles. In the GUI open Access > Federation > OAuth Authorization Server, then Local Traffic > Virtual Servers > Virtual Server List and read the Access Policy and profile settings. Then note the virtual server address and port and check which self IPs and firewall rules expose it to the internet or untrusted networks. Management (control plane) access is not the route. | | APM is licensed and provisioned | Run `tmsh list sys provision apm` and look for a level other than none, or open System > Resource Provisioning in the GUI. | ## Affected and fixed versions | Product line | Affected builds | First fixed build | | --- | --- | --- | | BIG-IP APM 21.1 | 21.1.0 up to before the hotfix | Hotfix-BIGIP-21.1.0.2.0.30.22-ENG | | BIG-IP APM 17.5 | 17.5.0 up to before the hotfix | Hotfix-BIGIP-17.5.1.9.0.160.12-ENG | | BIG-IP APM 17.1 | 17.1.0 up to before the hotfix | Hotfix-BIGIP-17.1.3.5.0.41.14-ENG | F5's CVE record lists the three lines above, each as affected from the first release shown up to the named hotfix, and names no other fixed build. The affected range for 21.1 starts at 21.1.0 and ends before the hotfix whose base is 21.1.0.2. Releases 17.5.1.9 and 17.1.3.5 without the hotfix are still affected. The hotfix names carry their base releases (17.5.1.9, 17.1.3.5 and 21.1.0.2). For 17.x, install that base release first, then the hotfix, following F5's usual hotfix procedure. For 21.1 the record states no separate step for 21.1.0 or 21.1.0.1; the hotfix name carries base 21.1.0.2, so plan the same order. Versions at End of Technical Support are not evaluated, so absence from the table does not prove an older line is safe. ## If the fix cannot go in yet - Install the hotfix for your line (versions table). It is an ENG hotfix image: install it to an inactive boot volume and reboot into it. On an HA pair, install on the standby unit first, reboot it, then fail over and repeat on the other unit. Failover drops connections that are not mirrored, and a single device is offline for the length of the reboot. This is F5's standard image-install method; K000162605 gives no separate downtime figure. - F5 lists one workaround: an iRule, available on request by opening a ticket with F5 support. F5 does not publish the iRule or say what it changes in function, so test it on a non-production unit first. - CISA says to apply the vendor-provided iRule as a temporary step to allow proactive forensic triage, and then install the final vendor patch as soon as possible. CISA marks forensic triage as required for this entry. - To shrink exposure until the hotfix is in, limit which networks can reach the virtual server that carries the OAuth profile (firewall rules, or a self IP port lockdown setting). Cost: clients outside the allowed networks can no longer reach that OAuth server. This is general hardening, not a step F5 lists for this CVE. ## What to look for - F5 and CISA publish no log strings, file paths or indicator addresses for this flaw. CISA marks forensic triage as required, so treat an internet-facing unpatched APM OAuth server as a candidate for triage under CISA's [BOD 26-04 implementation guidance](https://www.cisa.gov/news-events/directives/bod-26-04-implementation-guidance-prioritizing-security-updates-based-risk). - General checks on an APM unit you suspect (these are standard BIG-IP places to look, not indicators F5 published for this CVE): list crash files with `ls -lt /var/core`; search the traffic management log for a restart of the traffic manager with `grep -iE "tmm.*(core|SIGSEGV|restart)" /var/log/ltm`; read the APM log at `/var/log/apm` for OAuth requests from unfamiliar addresses; and compare the traffic manager's uptime with your change records using `bigstart status tmm`. A restart or a core file with no planned change on a unit that carries an OAuth Authorization Server is a reason to start forensic triage. ## Timeline | Date | Event | Source | | --- | --- | --- | | 20 Sep 2026 | CVE-2026-94127 is reserved by F5. | [source](https://cveawg.mitre.org/api/cve/CVE-2026-94127) | | 22 Sep 2026 | F5 publishes advisory K000162605 and the CVE record (public date 14:00 UTC). GitHub mirrors it as GHSA-qppv-6jrg-hxq4. F5 credits its own internal discovery. | [source](https://cveawg.mitre.org/api/cve/CVE-2026-94127) | | 22 Sep 2026 | NVD publishes the record with F5's CVSS 4.0 score of 9.3 and CVSS 3.1 score of 9.8. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-94127) | | 22 Sep 2026 | CISA adds the CVE to KEV with forensic triage required, ransomware use Unknown and a due date of 25 Sep 2026. The KEV note tells operators to apply the vendor-provided iRule as a temporary step, then install the final patch. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-94127) | The earliest dated mark of exploitation is CISA's, on 22 Sep 2026, the day of disclosure; exploitation may have begun earlier. ## Exploitation reported so far F5's advisory does not say it has seen exploitation, and no public report names the first attack, the actor or the method. ## Why static reachability settles it Exposure turns on the build and one configuration fact: an OAuth Authorization Server profile on a virtual server that an attacker can reach. Both can be read from the BIG-IP configuration. KEV tells you attacks occurred, but only the configuration read tells you whether this device can be hit. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-76460: Cisco Identity Services Engine](/ledger/cve-2026-76460/) - [CVE-2026-94545: Vercel Satori and Next.js](/ledger/cve-2026-94545/) - [CVE-2026-81963: Microsoft Windows Update Stack](/ledger/cve-2026-81963/) ## Sources - CVE-2026-94127: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-94127 (read 19 Nov 2026) - CVE-2026-94127: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-94127 (read 19 Nov 2026) - CVE-2026-94127: F5 advisory K000162605 (vendor advisory, as cited by the F5 CVE record): https://my.f5.com/manage/s/article/K000162605 (read 19 Nov 2026) - CVE-2026-94127: F5 CVE record on CVE Services (vendor advisory data: affected hotfix builds, workaround): https://cveawg.mitre.org/api/cve/CVE-2026-94127 (read 19 Nov 2026) - CVE-2026-94127: GitHub advisory GHSA-qppv-6jrg-hxq4: https://api.github.com/advisories?cve_id=CVE-2026-94127 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-94127/ --- # CVE-2026-76460: Cisco Identity Services Engine, is it exploitable where you run it Ledger entry, disclosed 16 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-76460 affects Cisco Identity Services Engine and was disclosed 16 Sep 2026 with a CVSS score of 10.0. Exposure needs these conditions: a node runs a release below the first fixed patch for its train, or ISE 3.0; the attacker can reach the node's API. CISA added it to the Known Exploited Vulnerabilities catalog on 16 Sep 2026. ## What the flaw is CVE-2026-76460 is an authentication bypass (CWE-648, incorrect use of privileged APIs per CISA) in an API of Cisco Identity Services Engine (ISE) and ISE Passive Identity Connector (ISE-PIC). Cisco says insufficient authentication control on an API endpoint lets an unauthenticated remote attacker send a crafted request and gain unauthorized access to the device. Cisco describes the bypass as going around the web-based management interface. ISE is the policy server that decides who joins the network, so control of it is control of network access. Cisco warns that on success an attacker may get command execution with root privileges. Cisco rates it 10.0 (CVSS 3.1 base score; network, low complexity, no privileges, scope changed, high impact on all three). It affects every ISE and ISE-PIC regardless of configuration, so there is no setting to check beyond the version and who can reach the node. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | a node runs a release below the first fixed patch for its train, or ISE 3.0 | Affected: an ISE or ISE-PIC node on a release below the first fixed patch for its train (3.1 Patch 12, 3.2 Patch 11, 3.3 Patch 12, 3.4 Patch 7, 3.5 Patch 4), or on 3.0, which is end of software maintenance. Run `show version` at the ISE CLI on each node. It prints the release and the installed patch level. In the admin portal, the patch management page lists installed patches. Compare every node with the versions table, since a node can lag its peers. Cisco's advisory names no ISE release earlier than 3.0 and its fixed table starts at 3.1, so a 2.x node is listed with no fix: treat it as affected and move it to a fixed 3.x train. Cisco says the flaw affects ISE and ISE-PIC regardless of device configuration, so no setting changes the result. | | the attacker can reach the node's API | Per Cisco's ISE API documentation, the ISE API Gateway receives API requests on HTTPS port 443, and from release 3.1 the ERS, Open API and monitoring (MnT) APIs are all routed through it. ERS clients use port 443 (9060 can also be used) and Open API clients use port 443; between nodes the gateway uses 9070 for Open APIs and 9443 for MnT. Cisco's advisory names no port for this CVE, and the flaw is in an API endpoint behind the management interface, so treat 443 on every node as the port to test, plus 9060 and 9070 where ERS and Open API are enabled (both are off by default). On each node run `show ports` at the ISE CLI to see what listens, and `show interface` for its addresses. Then list which networks can open those ports: test from a normal user VLAN and from any guest or internet-facing network, and read the firewall rules and infrastructure ACLs in front of the nodes. Cisco's mitigation is to allow only required management and control plane traffic to the nodes. | ## Affected and fixed versions | Release | Affected | First fixed release | | --- | --- | --- | | ISE 3.0 | All builds (end of software maintenance) | None: migrate to a supported release that has the fix | | ISE and ISE-PIC 3.1 | Earlier than 3.1 Patch 12 | 3.1 Patch 12 | | ISE and ISE-PIC 3.2 | Earlier than 3.2 Patch 11 | 3.2 Patch 11 | | ISE and ISE-PIC 3.3 | Earlier than 3.3 Patch 12 | 3.3 Patch 12 | | ISE and ISE-PIC 3.4 | Earlier than 3.4 Patch 7 | 3.4 Patch 7 | | ISE and ISE-PIC 3.5 | Earlier than 3.5 Patch 4 | 3.5 Patch 4 | Cisco's fixed-release table covers ISE and ISE-PIC together, so one patch level applies to both. The NVD record lists the highest affected ISE patches as 3.1 Patch 11, 3.2 Patch 10, 3.3 Patch 11, 3.4 Patch 6 and 3.5 Patch 3, and names ISE-PIC only at releases 3.4.0 and 3.5.0, the ISE-PIC releases it covers. Cisco's advisory is the source for the fixed patch levels. ## If the fix cannot go in yet - No workaround exists (Cisco). Upgrading to the fixed patch is the fix. - Cisco's mitigation: use infrastructure access control lists (iACLs) to allow only required management and control plane traffic to the ISE nodes. Cost: any admin station, integration or device that talks to ISE from a network you do not list loses access, so inventory who calls ISE first. It reduces who can reach the API but does not fix the flaw. - If you suspect compromise, Cisco strongly recommends re-imaging the affected nodes and restoring from a configuration backup if needed. Cost: node downtime and a rebuild, which also removes any attacker persistence. - Check for compromise now, not after patching. CISA's KEV entry marks forensic triage as required with a three-day due date and ties the required action to BOD 26-04. This is the same step as Cisco's re-image advice: if the access.log review finds anything, or you cannot rule out use before the patch, treat the node as compromised and re-image it. ## What to look for - Cisco's indicator of compromise is a review of access.log for suspicious usernames, on every node of a distributed deployment. Cisco gives no list of suspicious names. A username is suspicious when it is not an account you created: compare the names in the log with your administrators (Administration > System > Admin Access > Administrators) and your API and integration accounts, and treat any other name, a random-looking string or a placeholder-style name as a hit. This reading is our guidance, not Cisco's. - Cisco's example command is `admin#show logging application ise-kong/access.log | include dummyuser`, where dummyuser is a placeholder for a suspicious username. Any entry in the output may indicate malicious activity. - For more access.log files, collect a support bundle with include debug logs selected and shared key encryption, then decrypt it and read ./ise/logs/apigateway/access.log.gz. - After a root compromise, evidence may be hidden or removed on the node, so Cisco advises cross-checking network and firewall logs outside the device for unexpected uploads from the ISE node to external IP addresses and downloads from malicious IP addresses. Cisco publishes no indicator addresses. ## Timeline | Date | Event | Source | | --- | --- | --- | | 16 Sep 2026 | Cisco publishes advisory cisco-sa-ISE-ABP-VNSW7Tn5 version 1.0 (Final) at 16:00 GMT with fixed patches (bug ID CSCww39530). | [source](https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5) | | 16 Sep 2026 | NVD publishes the record the same day, with Cisco's CVSS 3.1 base score of 10.0. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-76460) | | 16 Sep 2026 | CISA adds the CVE to KEV with a federal due date of 19 Sep 2026, three days after the listing, and marks forensic triage as required (forensicTriage Yes): under BOD 26-04 an agency must check for compromise as well as apply the mitigation. Ransomware use is listed as Unknown. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-76460) | The earliest dated mark of exploitation is CISA's, on 16 Sep 2026, the day of disclosure; exploitation may have begun earlier. ## Exploitation reported so far Cisco's advisory says its PSIRT is aware of active exploitation and gives no date. It says the flaw was found during the resolution of a Cisco Technical Assistance Center support case. CISA's SSVC data in NVD records exploitation as active, automatable as yes and technical impact as total. ## Why attack-path analysis settles it Every unpatched ISE node is affected whatever its configuration, so the version only starts the question. What separates a node under attack from a node that is merely unpatched is whether anything hostile can open an HTTPS session to its management interface, and that depends on the firewall rules, infrastructure ACLs and segmentation in front of each node. Mapping the paths from user, guest and internet networks to each node settles that, and active exploitation makes it urgent. ## Related pages - [CVE-2026-81963: Microsoft Windows Update Stack](/ledger/cve-2026-81963/) - [CVE-2026-94127: F5 BIG-IP APM, BIG-IP APM OAuth server RCE](/ledger/cve-2026-94127/) - [CVE-2026-75650: Adobe Commerce and Magento Open Source](/ledger/cve-2026-75650/) ## Sources - CVE-2026-76460: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-76460 (read 19 Nov 2026) - CVE-2026-76460: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-76460 (read 19 Nov 2026) - CVE-2026-76460: Cisco security advisory cisco-sa-ISE-ABP-VNSW7Tn5 (affected products, indicators of compromise, mitigation, fixed releases): https://sec.cloudapps.cisco.com/security/center/content/CiscoSecurityAdvisory/cisco-sa-ISE-ABP-VNSW7Tn5 (read 19 Nov 2026) - CVE-2026-76460: Cisco ISE API documentation (API Gateway on port 443; ERS 9060; Open API 9070 and MnT 9443 between nodes): https://developer.cisco.com/docs/identity-services-engine/latest/ (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-76460/ --- # CVE-2026-81963: Microsoft Windows Update Stack, is it exploitable where you run it Ledger entry, disclosed 8 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-81963 affects Microsoft Windows Update Stack and was disclosed 8 Sep 2026 with a CVSS score of 7.8. Exposure needs these conditions: the host runs Windows 11 23H2, 24H2, 25H2 or 26H1, or Windows Server 2025, below the fixed build for its release; an attacker already runs code on the host as an authorized, low-privileged local user. ## What the flaw is CVE-2026-81963 is a link following flaw (CWE-59) in the Windows Update Stack. KEV also lists CWE-284. The stack resolves a file link improperly before it accesses a file, and Microsoft says an authorized local attacker can use that to gain SYSTEM privileges. No user interaction is needed. The flaw is local only. The attacker must already run code on the host as a low-privileged user, so it is an escalation step and not initial access. The CVSS 3.1 base score is 7.8 (local, low complexity, low privileges, high impact on confidentiality, integrity and availability). ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the host runs Windows 11 23H2, 24H2, 25H2 or 26H1, or Windows Server 2025, below the fixed build for its release | Run winver, or in PowerShell [System.Environment]::OSVersion.Version, or cmd /c ver, and compare the full build and revision with the table. Get-HotFix -Id KB5124008 (use the KB from the table) also works, but a later cumulative update contains the fix, so compare builds when the KB is absent. | | an attacker already runs code on the host as an authorized, low-privileged local user | This is a property of your environment (AV:L, PR:L). On a Windows 11 workstation every signed-in standard account is a local user, so the condition holds on every unpatched client. On Windows Server 2025 it depends on who can sign in: list them with net localgroup "Remote Desktop Users" and net localgroup Users, and review the Allow log on locally and Allow log on through Remote Desktop Services rights. Also consider any service or app on the host that runs untrusted code. | ## Affected and fixed versions | Windows release | Affected builds | First fixed build | Update (8 Sep 2026) | | --- | --- | --- | --- | | Windows 11 Version 23H2 (ARM64, x64) | Earlier than 10.0.22631.7582 | 10.0.22631.7582 | KB5122880 | | Windows 11 Version 24H2 (ARM64, x64) | Earlier than 10.0.26100.9445 | 10.0.26100.9445 | KB5124008 | | Windows 11 Version 25H2 (ARM64, x64) | Earlier than 10.0.26200.9445 | 10.0.26200.9445 | KB5124008 | | Windows 11 Version 26H1 (ARM64, x64) | Earlier than 10.0.28000.2954 | 10.0.28000.2954 | KB5124012 | | Windows Server 2025 (and Server Core) | Earlier than 10.0.26100.33438 | 10.0.26100.33438 | KB5122871 | Builds and KBs are from Microsoft's CVRF listing for this CVE (September 2026 release). Windows Server 2025 and Windows 11 24H2 share the 10.0.26100 base but are serviced separately: Microsoft lists 10.0.26100.33438 (KB5122871) for Server 2025 and 10.0.26100.9445 (KB5124008) for 24H2; 25H2 uses the 10.0.26200 base with 10.0.26200.9445 under the same KB5124008. Windows 10 and Windows Server 2012 to 2022 are not listed for this CVE. ## If the fix cannot go in yet - Install the September 2026 cumulative update, or a later one, for the release (see the versions table). Cost: a restart. This is the fix. - Microsoft lists no workaround. Two compensating controls reduce who can use the flaw. First, limit interactive and Remote Desktop logon to the accounts that need it, by removing groups from the Allow log on locally and Allow log on through Remote Desktop Services rights; cost: standard users and helpdesk staff lose access to those hosts. Second, enforce application control (WDAC or AppLocker) so that standard users cannot run unsigned or untrusted binaries; cost: a policy to build and maintain, and blocked tools until they are allowed. ## What to look for - Microsoft and CISA publish no indicator for this flaw, and CISA marks forensic triage as not required. Look for links that standard users create in folders the update stack reads or writes, such as C:\Windows\SoftwareDistribution and the servicing folders under C:\Windows. Sysmon does not log link creation as an event of its own, so search process creation (Sysmon event ID 1 or Security event 4688) for the commands that make links: `Get-WinEvent -FilterHashtable @{LogName='Microsoft-Windows-Sysmon/Operational'; Id=1} | Where-Object { $_.Message -match 'mklink\s+/[JD]|-ItemType\s+(Junction|SymbolicLink)' }`. In Microsoft Defender for Endpoint the same filter is `DeviceProcessEvents | where ProcessCommandLine matches regex @"mklink\s+/[JD]|-ItemType\s+(Junction|SymbolicLink)"` with the account name excluding SYSTEM. To find links already in place, run `Get-ChildItem C:\Windows\SoftwareDistribution -Recurse -Force -Attributes ReparsePoint` and inspect any result with `fsutil reparsepoint query `. Microsoft has not said which paths this CVE uses, so treat any such link as a lead, not proof. ## Timeline | Date | Event | Source | | --- | --- | --- | | 8 Sep 2026 | Microsoft publishes the advisory (revision 1.0) and the fix, marked Exploited: Yes and Publicly Disclosed: No. NVD publishes the record. CISA adds the CVE to KEV with a due date of 22 Sep 2026. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep) | | 11 Sep 2026 | Revision 1.1 updates an acknowledgement (informational change only). | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep) | | 15 Sep 2026 | Revision 1.2 updates an acknowledgement (informational change only). | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep) | | 6 Oct 2026 | Revision 1.3 updates an acknowledgement (informational change only). | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep) | | 7 Oct 2026 | Revision 1.4 updates an acknowledgement (informational change only). The credits now name Zhang WangJunJie with Hillstone Networks, Romain Deperne, haowei yan and MSTIC. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep) | Exploited on the day of disclosure, 8 Sep 2026. ## Exploitation reported so far Microsoft marked the flaw as exploited on the day it published the fix, 8 Sep 2026, and CISA added it to KEV the same day. KEV lists ransomware campaign use as Unknown. Microsoft credits Zhang WangJunJie with Hillstone Networks, Romain Deperne, haowei yan and the Microsoft Threat Intelligence Center (MSTIC) as reporters. No attack details have been published. ## Why attack-path analysis settles it Exposure needs an unpatched Windows 11 or Server 2025 build and an attacker who already runs code locally on that host. The build check settles the first. As a local escalation step, an attack-path view of how an attacker reaches local code settles real priority, and KEV shows it is already used. ## Related pages - [CVE-2026-75650: Adobe Commerce and Magento Open Source](/ledger/cve-2026-75650/) - [CVE-2026-76460: Cisco Identity Services Engine](/ledger/cve-2026-76460/) - [CVE-2026-87491: Google Chromium V8](/ledger/cve-2026-87491/) ## Sources - CVE-2026-81963: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-81963 (read 19 Nov 2026) - CVE-2026-81963: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-81963 (read 19 Nov 2026) - CVE-2026-81963: Microsoft advisory (MSRC CVRF, September 2026 release: fixed builds, KBs, Exploited: Yes, revision history, acknowledgments): https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Sep (read 19 Nov 2026) - CVE-2026-81963: Microsoft advisory (MSRC API record): https://api.msrc.microsoft.com/sug/v2.0/en-US/vulnerability/CVE-2026-81963 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-81963/ --- # CVE-2026-75650: Adobe Commerce and Magento Open Source, is it exploitable where you run it Ledger entry, disclosed 7 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-75650 affects Adobe Commerce and Magento Open Source and was disclosed 7 Sep 2026 with a CVSS score of 10.0. Exposure needs these conditions: the store runs Adobe Commerce, B2B or Magento Open Source through 2026-aug; hotfix VULN-39341 is not installed. CISA added it to the Known Exploited Vulnerabilities catalog on 8 Sep 2026. Exploited 3 days before disclosure. ## What the flaw is Adobe Commerce and Magento fill email and page templates by replacing directives such as variable and block tags. A flaw in how the template engine handles nested directives lets an unauthenticated visitor smuggle a directive through guest cart address fields. Before the patch, a smuggled block directive created any PHP class, not only block classes. That gives the attacker arbitrary object creation, and from there code execution as the web user (CWE-1336, template engine injection). No user interaction is needed and the scope is changed, so Adobe rates it 10.0 (CVSS 3.1 base score). Sansec and Fastly describe an attack in stages: first get attacker-controlled PHP text onto the disk, for example through a log or error report that stores a request unescaped, then trigger the template flaw through Magento's standard failed payment email (the 'Payment Transaction Failed Reminder' template, present in a standard install) so that a gadget includes that file. Sansec says the attack can also succeed when email delivery fails, so turning off outbound mail is no fix, and that more than one way to write the text exists, so closing one write path does not stop it. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the store runs Adobe Commerce, B2B or Magento Open Source through 2026-aug | Affected: Adobe Commerce 2.4.4 to 2.4.9, Commerce B2B 1.3.3 to 1.5.3, or Magento Open Source 2.4.6 to 2.4.9 (Sansec also reports Open Source 2.4.4 and 2.4.5) at any patch level through the 2026-aug releases, or an older branch that has no Adobe fix. On the server run bin/magento --version. Or run composer show magento/product-community-edition (Open Source) or composer show magento/product-enterprise-edition (Commerce), and composer show magento/extension-b2b for B2B. The admin footer also shows the version. Adobe's record lists Open Source 2.4.6 to 2.4.9; Sansec says the hotfix was tested on Open Source 2.4.4 to 2.4.9 and that those releases are exposed, so treat 2.4.4 and 2.4.5 as affected. Branches before 2.4.4 are affected and get no Adobe fix. | | Hotfix VULN-39341 is not installed | From the Magento root run vendor/bin/magento-patches -n status \| grep "39341\\|Status". Sansec gives this command from Adobe's bulletin; the line for 39341 must show Applied. Any other result, or no line for 39341, means the store is unpatched. | | The storefront is reachable by unauthenticated visitors from the internet, including guest cart and GraphQL endpoints | From a host outside your network run curl -s -o /dev/null -w '%{http_code}\n' -H 'Content-Type: application/json' -d '{"query":"{ __typename }"}' https://shop.example/graphql. A 200 means the GraphQL endpoint answers without login. Also confirm guest checkout is enabled in the admin under Stores > Configuration > Sales > Checkout > Checkout Options > Allow Guest Checkout. | | The web user can write to places Magento later includes (var/log, var/report, pub/media) | Run sudo -u test -w var/report && echo writable, and the same for var/log and pub/media. These are writable on a standard install, so assume yes unless you have hardened the layout. | ## Affected and fixed versions | Product line | Affected | Fix | | --- | --- | --- | | Adobe Commerce 2.4.9, 2.4.8, 2.4.7, 2.4.6, 2.4.5, 2.4.4 | All releases and patch levels through 2.4.x-2026-aug | Hotfix VULN-39341 (composer patch) | | Adobe Commerce B2B 1.5.3, 1.5.2, 1.4.2, 1.3.4, 1.3.3 | All releases and patch levels through 1.x.y-2026-aug | Hotfix VULN-39341 (composer patch) | | Magento Open Source 2.4.9, 2.4.8, 2.4.7, 2.4.6 | All releases and patch levels through 2.4.x-2026-aug (Adobe's record lists 2.4.6 to 2.4.9; Sansec says 2.4.5 and 2.4.4 are affected and tested the hotfix there) | Hotfix VULN-39341 (composer patch) | | Branches 2.2, 2.3, 2.4.0 to 2.4.3 (end of support) | Affected | No Adobe fix. Scandiweb published backports for 41 older releases, 2.2.0 to 2.4.3-p3 (third-party patches, test on staging first) | The fix is a hotfix, not a new release; apply and verify it as described under the mitigation steps. Adobe tested it on the 2026-aug releases; older patch levels inside those branches are affected and the patch is unverified there (Sansec). In Adobe's CNA data in NVD the fixed state reads 'Hotfix for CVE-2026-7565', one digit short of this CVE number and apparently a typographical slip. ## If the fix cannot go in yet - Apply Adobe's hotfix VULN-39341 (APSB26-146): download VULN-39341-composer-patches.zip from repo.magento.com and apply it as a composer patch, then run the status command from the checks table; the line for 39341 must show Applied. Cost: a composer patch run and a deploy window; no new release number is involved. - Compensating control: a WAF rule. Fastly offers a virtual patch for its Next-Gen WAF covering each stage, and Sansec Shield blocks the variants it has seen. Both vendors say to install the Adobe patch anyway, since operators changed payloads several times a day from 4 Sep. Cost: possible false positives on admin template updates. - Moving sessions to Redis or the database does not stop the attack (Sansec). Disabling mail does not stop it either. - Unsupported branches (2.2, 2.3, 2.4.0 to 2.4.3) get no Adobe patch. Upgrade to a supported release. Scandiweb publishes third-party backports for older releases; test them on a staging copy first. - CISA's KEV entry gives the required action: apply mitigations in accordance with vendor instructions, follow CISA's Forensics Triage Requirements, and discontinue use of the product if mitigations are unavailable. The entry also points to the BOD 26-04 patching guidelines. ## What to look for - Host evidence (Sansec). A strong sign is a background process disguised as [kworker/u:8:0], fc-cache or chronyd, run by the web user. Run crontab -l | grep -i gvfsd; ls -la ~/.local/share/.gvfsd/ ~/.cache/fontconfig/fc-cache /tmp/.kw_* /tmp/.cache_* /tmp/.gvfsd-* /tmp/.fc-*/fc-cache /tmp/fc-cache /tmp/.chrony-*/chronyd 2>/dev/null; ps -eo pid,comm,args | grep -iE 'kworker|fc-cache|chronyd'. A chronyd or fc-cache process whose parent is PID 1, or a cron line such as 13,43 * * * * for fc-cache, is suspect. An empty crontab is not proof of a clean host, because one build can relaunch with no cron entry. Check the cron spool file itself, not only crontab -l. - Logs and files (Sansec): grep -r 'crontab command not allowed' /var/log/ ; grep -ril 'x_trace_' var/report/ ; find -L pub/media -type f -name '*.php'. A PHP file under pub/media/catalog/product/cache/ss_<10 hex>/ named sync_<10 hex>.php is a web shell. A modified vendor/magento/framework/App/View.php that loads remote code is another sign. Bursts of failed payment emails in the mail log deserve a look; ordinary declined payments send the same message. - Network indicators (Sansec, defanged): C2 99[.]84[.]67[.]186:443 (WebSocket over TLS); NTP-shaped UDP port 123 beacons to ntp[.]timesync[.]to and ntp[.]timesysnc[.]net, both resolving to 185[.]157[.]160[.]251 on 7 Sep 2026, with fallbacks ntp[.]synctime[.]to and ntp[.]syncstime[.]to; malware download host 209[.]141[.]43[.]95; second toolkit C2 213[.]111[.]146[.]27 on ports 80, 443 and 8443. A real NTP client sends one datagram per poll; the implant sends nine 48-byte datagrams about 10 ms apart every 60 seconds. - Attacker source addresses Sansec lists: 88[.]216[.]72[.]181, 88[.]216[.]72[.]182, 182[.]182[.]152[.]48, 76[.]31[.]99[.]207, 209[.]73[.]130[.]148, 77[.]239[.]124[.]107, 98[.]224[.]189[.]28. SHA-256 of the fc-cache family and other samples are in Sansec's article. Scan with Sansec eComscan. If a suspicious process turned up, rotate Magento credentials. Patching closes the hole but does not clean a store that was already hit. - After exposure (Adobe recommendation, per Sansec): rotate the encryption key and every credential it protected: admin passwords, REST, SOAP and GraphQL integration tokens, OAuth client secrets, payment gateway API credentials, database credentials, SSH and deploy keys, and third-party extension API keys. Rotate at the source, not only in Magento, because rotating the encryption key alone does not invalidate anything an attacker already read. Cost: every integration needs a new credential. Rotation does not replace the hotfix. ## Timeline | Date | Event | Source | | --- | --- | --- | | 4 Sep 2026 | Sansec finds the campaign at 22:40 UTC; attacks started this day. The first victim ran 2.4.6-p15 with the July and August 2026 patches. | [source](https://sansec.io/research/stylesmuggler-0day) | | 5 Sep 2026 | Sansec publishes StyleSmuggler and its Shield rules go live in the early morning. | [source](https://sansec.io/research/stylesmuggler-0day) | | 7 Sep 2026 | Adobe publishes bulletin APSB26-146 (priority 1, hotfix VULN-39341; Sansec gives 20:20 UTC) and NVD publishes the record from Adobe's CNA data at 21:17 UTC: CVSS 3.1 10.0, CWE-1336. | [source](https://helpx.adobe.com/security/products/magento/apsb26-146.html) | | 8 Sep 2026 | CISA adds the CVE to KEV, due date 11 Sep 2026, forensic triage Yes. CISA's SSVC entry in NVD: exploitation active, automatable yes, technical impact total. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-75650) | | 16 Sep 2026 | Fastly publishes its analysis, a Nuclei template and a virtual patch. It says Adobe's patch addresses portions of the vulnerability chain. | [source](https://www.fastly.com/blog/what-is-style-smuggler-cve-2026-75650) | Exploited 3 days before disclosure: disclosed 7 Sep 2026, first exploitation reported 4 Sep 2026. ## Exploitation reported so far Sansec reproduced the chain on clean Magento Open Source 2.4.7, 2.4.8 and 2.4.9. Two operators were seen: one that installs a Rust backdoor with NTP-shaped command traffic, and a second that drops a web shell in the product image cache. A later store compromise used gs-netcat and the WraithC2 remote access tool. Sansec says it has no sign the backdoor was weaponized further. ## Why known-exploited lists settle it Attackers used this flaw three days before the hotfix existed, it needs no login, and Sansec shows a store with the July and August patches was hit. Known-exploited status means two actions on every internet-facing store: confirm the hotfix is applied, and investigate for compromise, because the hotfix closes the hole but does not clean a store that was already hit. The version number alone does not settle it, since the fix is a hotfix, not a release. See [known-exploited lists](/methods/known-exploited/) for what the method proves and misses. ## Related pages - [CVE-2026-87491: Google Chromium V8](/ledger/cve-2026-87491/) - [CVE-2026-81963: Microsoft Windows Update Stack](/ledger/cve-2026-81963/) - [CVE-2026-84869: ConnectWise ScreenConnect](/ledger/cve-2026-84869/) ## Sources - CVE-2026-75650: NVD record carrying Adobe's advisory data for APSB26-146 (affected versions, hotfix, CVSS): https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75650 (read 19 Nov 2026) - CVE-2026-75650: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-75650 (read 19 Nov 2026) - CVE-2026-75650: Sansec research: StyleSmuggler 0-day (hotfix details, versions tested, indicators, timeline): https://sansec.io/research/stylesmuggler-0day (read 19 Nov 2026) - CVE-2026-75650: Fastly report: What is Style Smuggler (16 Sep 2026): https://www.fastly.com/blog/what-is-style-smuggler-cve-2026-75650 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-75650/ --- # CVE-2026-87491: Google Chromium V8, is it exploitable where you run it Ledger entry, disclosed 4 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-87491 affects Google Chromium V8 and was disclosed 4 Sep 2026 with a CVSS score of 8.8. Exposure needs these conditions: Chrome runs a build below the fixed build in the versions table, or was updated but not relaunched; the user can be made to open attacker-controlled content (a link, an ad or a compromised site). ## What the flaw is This flaw is an out-of-bounds write in V8, the JavaScript and WebAssembly engine that Chrome and every Chromium-based browser share: the engine writes outside a memory buffer it should stay within. A remote attacker can use it through a crafted HTML page to run code inside the browser's sandbox. The user needs to load the page, and nothing else (CVSS: network, no privileges, user interaction required). The code runs inside the renderer sandbox, so it does not by itself give control of the machine. Attackers usually chain a flaw like this with a separate sandbox escape. Google has not published details of the chain. Chromium's release note rates the bug Medium, while the CVSS 3.1 score is 8.8 (High), calculated by CISA as an additional scorer in NVD. Google gives no reason for the Medium label. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | Chrome runs a build below the fixed build in the versions table, or was updated but not relaunched | Chrome applies an update only after a relaunch, and the same holds for any other Chromium-based browser or embedded Chromium app whose Chromium is older than its fixed build. Open chrome://version, or Menu > Help > About Google Chrome, which also starts the update. At scale: Windows `(Get-Item "$env:ProgramFiles\Google\Chrome\Application\chrome.exe").VersionInfo.ProductVersion`; Linux `google-chrome --version`; macOS `/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --version`. In Chrome Browser Cloud Management, open the Browsers list and read the version column. A pending update shows a Relaunch button on chrome://settings/help; a device that was never restarted still runs the old build even if the new files are on disk. To close the gap on a fleet, set the relaunch policies listed under the mitigation. | | the user can be made to open attacker-controlled content (a link, an ad or a compromised site) | Review whether managed browsers let users visit the open web. Web filtering and isolation policies change the odds but do not remove the condition. | | another Chromium-based browser (Edge, Brave, Opera and others) is older than the build that carries the fix | Edge: Settings > About Microsoft Edge (edge://version); the fixed Edge builds are in the versions table. For other browsers, ask the vendor which release ingests the fix; chrome://version shows the Chromium build, and 153.0.8010.36 is the fixed build on the 153 line. | | Embedded Chromium or V8 runs untrusted web content (Electron apps, Chromium Embedded Framework, WebView-based tools) | For Electron, read the bundled Chromium with `process.versions.chrome` in the app console. Electron publishes no fixed version for this CVE. As of 19 Nov 2026 Electron's releases data lists no release that bundles Chromium 153; the newest, 44.11.5 (18 Nov 2026), bundles 152.0.7977.169, and neither Google nor Electron says that build carries the fix. So an Electron app has no confirmed fixed version: ask the app's vendor which release does, and until then keep the app away from untrusted pages (load only your own content, open external links in the default browser). An app that bundles Chromium 153.0.8010.36 or later is fixed. Chrome for Android 153.0.8010.36 carries the same security fixes as the desktop release, according to Google's release note of 8 Sep 2026. That note does not mention Android WebView, so check the WebView version separately on devices that use it. | ## Affected and fixed versions | Product | Affected builds | First fixed build | | --- | --- | --- | | Google Chrome, Linux | Earlier than 153.0.8010.36 | 153.0.8010.36 | | Google Chrome, Windows and Mac | Earlier than 153.0.8010.36/.37 | 153.0.8010.36/.37 | | Google Chrome for Android | Earlier than 153.0.8010.36 | 153.0.8010.36 (same fixes as desktop) | | Google Chrome 152 (Stable 152.0.7977.82/.83, Extended Stable 152.0.7977.120) | Google names no 152 build with the fix | Move to 153.0.8010.36 or later | | Microsoft Edge (Chromium-based) | Builds before the Microsoft fix | 152.0.4191.66 (4 Sep 2026) on the 152 line, or 153.0.4234.32 (10 Sep 2026) and later | NVD gives the Chrome range as below 153.0.8010.36. Google's 3 Sep 2026 Stable post (152.0.7977.82/.83) and its 9 Sep 2026 Extended Stable post (152.0.7977.120) do not list this CVE; only the 8 Sep 2026 post for Chrome 153 does. Microsoft's Edge release notes of 4 Sep 2026 say Edge 152.0.4191.66 contains a fix for this CVE; Edge 153.0.4234.32 followed on 10 Sep 2026. ## If the fix cannot go in yet - Where users defer updates, enforce a relaunch deadline with Chrome policy: set RelaunchNotification to 2 (required) and RelaunchNotificationPeriod to the number of milliseconds users get before Chrome relaunches itself (the default is 604800000, 7 days; 3600000 is 1 hour). Cost: users lose the option to defer, and open tabs reload on the forced relaunch. - Google lists no workaround or settings change for this flaw. Until the update lands, CISA's KEV action is to apply the vendor's mitigations, or stop using the product if none exist. Limiting exposure to untrusted sites for unpatched browsers (web filtering, isolation) reduces risk but is our suggestion, not Google's. Cost: blocked sites. ## What to look for - Google, Microsoft and CISA publish no indicators of compromise, log strings or file paths for this flaw, and the Chromium issue tracker holds no public detail. The vendor guidance is to update. - Endpoint inventory (Intune, Jamf, Chrome Browser Cloud Management or similar) can list browsers below the fixed version so you can force an update and relaunch. ## Timeline | Date | Event | Source | | --- | --- | --- | | 6 Aug 2026 | Jihyeon Jeong (Compsec Lab, Seoul National University) reports the bug, Chromium issue 543557673. Reward $2,500. | [source](https://chromereleases.googleblog.com/2026/09/stable-channel-update-for-desktop_0808145027.html) | | 4 Sep 2026 | Microsoft releases Edge 152.0.4191.66. Its Edge release notes of that date say the Chromium team reported that CVE-2026-87491 has an exploit in the wild and that the update contains a fix for it. This is the earliest dated public report of exploitation, and the first public mention of the CVE. | [source](https://learn.microsoft.com/en-us/deployedge/microsoft-edge-relnotes-security) | | 8 Sep 2026 | Chrome 153.0.8010.36 (Linux) and 153.0.8010.36/.37 (Windows, Mac) reach the stable channel. Google's release note says it is aware that an exploit for CVE-2026-87491 exists in the wild, and rates the bug Medium. Chrome 153 for Android follows the same day. | [source](https://chromereleases.googleblog.com/2026/09/stable-channel-update-for-desktop_0808145027.html) | | 9 Sep 2026 | NVD publishes the record and CISA adds the CVE to KEV, due date 23 Sep 2026. CISA's SSVC entry in NVD (10 Sep) marks exploitation as active, not automatable, with total technical impact. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-87491) | | 10 Sep 2026 | Microsoft releases Edge 153.0.4234.32. | [source](https://learn.microsoft.com/en-us/deployedge/microsoft-edge-relnotes-security) | | 11 Sep 2026 | Microsoft lists the CVE in its Security Update Guide, noting Edge ingests the Chromium fix and repeating Google's exploit statement. | [source](https://api.msrc.microsoft.com/sug/v2.0/en-US/vulnerability/CVE-2026-87491) | Exploited on the day of disclosure, 4 Sep 2026. ## Exploitation reported so far Microsoft's Edge note of 4 Sep 2026, the first public mention of the CVE, already says the Chromium team reported an exploit in the wild. Google's own note followed on 8 Sep 2026. Neither vendor names an actor, target or date of first use. ## Why agent investigation settles it Exposure is per endpoint: the browser build, whether it was relaunched, and which other Chromium-based apps are installed. An agent or endpoint tool that reads those on each machine settles it, while network reachability tells you nothing because the attacker comes through the user's browser. KEV tells you to treat every unpatched endpoint as a target. ## Related pages - [CVE-2026-84869: ConnectWise ScreenConnect](/ledger/cve-2026-84869/) - [CVE-2026-75650: Adobe Commerce and Magento Open Source](/ledger/cve-2026-75650/) - [CVE-2026-82329: JFrog Artifactory](/ledger/cve-2026-82329/) ## Sources - CVE-2026-87491: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-87491 (read 19 Nov 2026) - CVE-2026-87491: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-87491 (read 19 Nov 2026) - CVE-2026-87491: vendor release note (Chrome Releases, Stable Channel Update for Desktop, 8 Sep 2026) advisory: https://chromereleases.googleblog.com/2026/09/stable-channel-update-for-desktop_0808145027.html (read 19 Nov 2026) - CVE-2026-87491: Chrome Releases blog feed copy of the same post: https://chromereleases.googleblog.com/feeds/posts/default?alt=json&q=CVE-2026-87491 (read 19 Nov 2026) - CVE-2026-87491: Chrome Releases, Stable Channel Update for Desktop, 3 Sep 2026 (152.0.7977.82/.83): https://chromereleases.googleblog.com/2026/09/stable-channel-update-for-desktop_01882797386.html (read 19 Nov 2026) - CVE-2026-87491: Microsoft Security Update Guide record (Edge fixed build): https://api.msrc.microsoft.com/sug/v2.0/en-US/vulnerability/CVE-2026-87491 (read 19 Nov 2026) - CVE-2026-87491: Microsoft Edge security release notes: https://learn.microsoft.com/en-us/deployedge/microsoft-edge-relnotes-security (read 19 Nov 2026) - CVE-2026-87491: Electron releases data (bundled Chromium per release): https://releases.electronjs.org/releases.json (read 19 Nov 2026) - CVE-2026-87491: Chrome Releases, Chrome for Android Update, 8 Sep 2026 (153.0.8010.36): https://chromereleases.googleblog.com/2026/09/chrome-for-android-update_01729269728.html (read 19 Nov 2026) - CVE-2026-87491: Chrome Releases, Extended Stable Update for Desktop, 9 Sep 2026 (152.0.7977.120): https://chromereleases.googleblog.com/feeds/posts/default?alt=json&published-min=2026-09-09T00:00:00&published-max=2026-09-10T00:00:00 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-87491/ --- # CVE-2026-84869: ConnectWise ScreenConnect, is it exploitable where you run it Ledger entry, disclosed 3 Sep 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-84869 affects ConnectWise ScreenConnect and was disclosed 3 Sep 2026 with a CVSS score of 9.9. Exposure needs these conditions: an endpoint runs a ScreenConnect client older than 26.6.5; a role holds TransferFiles and an attacker controls a session peer. CISA added it to the Known Exploited Vulnerabilities catalog on 11 Sep 2026. ## What the flaw is Older ScreenConnect clients in Support and Access sessions could process file-transfer actions through an active session without proper authorization or Host confirmation (CWE-862 and CWE-269). Under certain circumstances, files could be transferred to and run on the Host client system, including through elevated execution. The server is not affected. The exposed systems are the endpoints and technician workstations that run the client. ConnectWise says it cannot give more detail because of the sensitive nature of the flaw. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | an endpoint runs a ScreenConnect client older than 26.6.5 | On a Windows host read the client file version: `(Get-Item "C:\Program Files (x86)\ScreenConnect Client (*)\ScreenConnect.ClientService.exe").VersionInfo.FileVersion`. At fleet scale, run the same query through your RMM, EDR or an Intune remediation script and collect every result, then list the hosts below 26.6.5. Cloud and on-premises are both in scope; Cloud servers were updated by ConnectWise but clients still need a reinstall. The server is not affected, and updating the server does not update clients. ConnectWise documents no version command for macOS or Linux clients. Read the version from your MDM or package inventory, and where none shows it, reinstall the client from the updated server, which ConnectWise says is required for host clients. | | a role holds TransferFiles and an attacker controls a session peer | Administration > Security > Roles: edit each role, review each session group with permissions (shown in bold), and check Scoped Permissions for TransferFiles (TransferFilesInSession on legacy versions). List the users in those roles; any one of them, or an attacker holding their account, can start a session. The peer can also be a Guest who joins a session a technician then opens. In Administration > Audit log, look for accounts or Guests you do not recognize, including blank-named guest sessions, and confirm MFA is on for technician accounts. | ## Affected and fixed versions | Deployment | Affected | Fixed | | --- | --- | --- | | On-premises ScreenConnect | Versions prior to 26.6.5 (clients) | 26.6.5 or later; source versions must be 25.4 or later to upgrade | | Cloud ScreenConnect | Clients prior to 26.6.5 | Servers updated by ConnectWise; reinstall host clients and update access agents | | Automate-integrated on-premises | Versions prior to 26.6.5 | 26.6.5 via Automate Product Updates (active Automate Assurance subscription) | NVD's CPE range ends before 26.6.5.9742, a more specific form of the same release number; the vendor bulletin states 26.6.5. Treat any client below 26.6.5 as affected. Access agents must also be updated after upgrading. ## If the fix cannot go in yet - Upgrade to 26.6.5 or later (primary fix). If your license is out of maintenance, renew or upgrade it first. Cost: an upgrade window and reinstalling host clients. - Interim, per ConnectWise: in Administration > Security > Roles, deselect TransferFiles (TransferFilesInSession on legacy versions) for every session group in every role, then save. Cost: technicians can no longer transfer files in sessions. It reduces exposure and does not replace the update. - After patching: review users and remove any not recognized, review roles and permissions, change passwords and enable MFA. - If you suspect compromise: isolate the technician workstations and endpoints that ran a client older than 26.6.5, back them up for analysis, and keep them offline until investigated, rebuilt and patched. The server is not vulnerable to this CVE, so server isolation applies only if you also suspect the server, for example through the third-party extension campaign that ConnectWise describes for on-premises servers. Reset passwords, review the audit log and force all technicians to sign back in (ConnectWise security alert checklist). ## What to look for - Server side, the CVE-specific check (Huntress, 16 Nov 2026): a file transfer by a Guest leaves RanFiles (EventType 85) or SentFiles (EventType 86) events in the server's audit log and in App_Data\Session.db (work on a copy), sent by a guest connection (ProcessType = 2). Look for repeated requests for update.bat or update_<8 hex>.bat from blank-named sessions, marker.txt, Run-Lateral.bat, svc_payload.cmd, and any file path that contains ../ or a Startup folder path. Then check whether a Host connection (ProcessType = 1) was open at that moment: that overlap is the exposure. - On technician workstations (Huntress): ScreenConnect.WindowsClient.exe, or your branded client, running .bat, .cmd, .ps1, .vbs, .js or .txt files from a Documents\...\Temp folder; unexpected files in %APPDATA%\Microsoft\Windows\Start Menu\Programs\Startup, especially Sync.bat; folders %LOCALAPPDATA%\wclient_<8 hex> and files %TEMP%\wupdate_<8 hex>.msi; WindowsUpdate.WindowsClient.exe or any client whose command line points at a relay you do not own, often on port 8041. - Indicators from Huntress's 12 Nov 2026 case: payload host and relay masterdarky[.]com (resolving to 194[.]15[.]36[.]33), installer WindowsUpdate.msi, SHA-256 of the rogue client f8871b1561d5c77c297378a7f49c84de52a01ff297395544f0e95d13ef972841, first-stage file update_224ac089.bat with argument xhsdf_wrwerhjf_shhdga, and guest source addresses 91[.]224[.]92[.]167, 193[.]26[.]115[.]63, 57[.]128[.]131[.]66, 91[.]224[.]92[.]83 and 107[.]148[.]223[.]37. - Separate from this CVE: Huntress's Aug and Sep 2026 reports of rogue ScreenConnect clients spawning wscript.exe to run 1.vbs to 4.vbs, with a WindowsServiceHost Run key, describe different intrusions that Huntress did not tie to this CVE. Treat those as general hunting indicators. - ConnectWise has published no indicators specific to this CVE. For its Nov 2026 third-party extension campaign it advises reviewing authentication, extension installation, administrative, Run Command and Toolbox records, revoking active web sessions, and investigating endpoints with EDR rather than ScreenConnect. ## Timeline | Date | Event | Source | | --- | --- | --- | | 3 Sep 2026 | ConnectWise publishes the interim Guest File Transfer advisory with the TransferFiles mitigation. | [source](https://www.connectwise.com/company/trust/advisories) | | 8 Sep 2026 | ConnectWise releases 26.6.5 and the bulletin for CVE-2026-84869, and NVD publishes the CVE. Huntress updates its Aug-Sep post on rogue ScreenConnect clients to say ConnectWise has fixed CVE-2026-84869; the post does not say its August incidents used the CVE. | [source](https://www.connectwise.com/company/trust/security-bulletins/2026-09-08-screenconnect-bulletin) | | 9 Sep 2026 | ConnectWise updates the interim advisory to point to the security bulletin. | [source](https://www.connectwise.com/company/trust/advisories) | | 10 Sep 2026 | CISA's SSVC data in the NVD record marks exploitation as active. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-84869) | | 11 Sep 2026 | CISA adds the CVE to KEV, due date 14 Sep 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-84869) | | 13 Nov 2026 | ConnectWise tells all partners to verify every technician client runs 26.6.5 or later and to disable listed third-party extensions. | [source](https://www.connectwise.com/company/trust/advisories) | | 16 Nov 2026 | ConnectWise reports its investigation complete and repeats the client version requirement. | [source](https://www.connectwise.com/company/trust/advisories) | | 16 Nov 2026 | Huntress publishes its confirmation of in-the-wild exploitation of CVE-2026-84869: a guest-pushed batch file ran on an MSP technician's workstation on 12 Nov 2026. It dates the first logged file push to a technician client to 19 Oct 2026 and a count of 688,000 guest run-file requests to the period from 20 Oct 2026. | [source](https://www.huntress.com/blog/cve-2026-84869-screenconnect-technician-exploitation) | The earliest dated mark of exploitation is CISA's, on 10 Sep 2026, 7 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far ConnectWise rates the bulletin Priority 1 High, its label for flaws being targeted or at higher risk of being targeted, and has published no date of first exploitation. The 10 Sep 2026 date is CISA's mark, not an observed attack, so the days-from-disclosure figure counts from ConnectWise's 3 Sep advisory to that mark; the first attack that an outside observer dated is Huntress's 19 Oct 2026 near-miss. Huntress says the 12 Nov 2026 batch file installed a rogue ScreenConnect client within about 9 seconds. After the 19 Oct Run-Lateral.bat push, which Huntress counts as a near-miss and not an exposure, pushes while a technician was connected followed on 23, 24 and 27 Oct 2026, including a path-traversal attempt aimed at a Startup folder. In Nov 2026 ConnectWise said third-party researchers observed a toolkit (SCKit) built to spread from infected endpoints to connecting technician workstations, and that client versions 26.6.5 and later block that spread technique. Huntress says SCKit appears to use a vulnerable third-party server extension and not this CVE directly. ## Why agent investigation settles it Exposure lives in the client on each endpoint, not the server, so the server version settles nothing. What decides it is each client's version together with which roles can transfer files and whether the session audit log shows a peer you do not recognize. That takes a client inventory plus an audit-log investigation. ## Related pages - [CVE-2026-82329: JFrog Artifactory](/ledger/cve-2026-82329/) - [CVE-2026-87491: Google Chromium V8](/ledger/cve-2026-87491/) - [CVE-2026-80521: Linux kernel](/ledger/cve-2026-80521/) ## Sources - CVE-2026-84869: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-84869 (read 19 Nov 2026) - CVE-2026-84869: CISA KEV entry (KEV feed 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-84869 (read 19 Nov 2026) - CVE-2026-84869: vendor advisory (ConnectWise security bulletin): https://www.connectwise.com/company/trust/security-bulletins/2026-09-08-screenconnect-bulletin (read 19 Nov 2026) - CVE-2026-84869: vendor advisory (ConnectWise disclosures repo): https://github.com/ConnectWise-Advisories/Disclosures/tree/main/CVE-2026-84869 (read 19 Nov 2026) - CVE-2026-84869: vendor advisories page (interim advisory 3 Sep and Nov 2026 updates): https://www.connectwise.com/company/trust/advisories (read 19 Nov 2026) - CVE-2026-84869: Huntress report on rogue ScreenConnect installations (updated 8 Sep 2026): https://www.huntress.com/blog/rogue-screenconnect-installations (read 19 Nov 2026) - CVE-2026-84869: Huntress report, CVE-2026-84869 exploited against MSP technician accounts, published 16 Nov 2026: https://www.huntress.com/blog/cve-2026-84869-screenconnect-technician-exploitation (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-84869/ --- # CVE-2026-82329: JFrog Artifactory, is it exploitable where you run it Ledger entry, disclosed 28 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-82329 affects JFrog Artifactory and was disclosed 28 Aug 2026 with a CVSS score of 9.8. Exposure needs self-hosted Artifactory (not JFrog Cloud) on a release below the first fixed build for its train. CISA added it to the Known Exploited Vulnerabilities catalog on 2 Sep 2026. Exploited 4 days after disclosure. ## What the flaw is The Access service in front of Artifactory builds a list of join keys it will trust when a node asks to register. It splits a configuration string on commas without dropping empty entries. On a default install that string is empty, so an empty key ends up on the trusted list (Fastly). Because the empty key's signing secret is fixed and derived from nothing, anyone can forge a token that the service accepts as a valid join request. The unauthenticated join endpoint then returns an admin-scoped token. JFrog classes it as improper authentication (CWE-287); the CNA score is 9.8. The attacker gains full control of the platform: users, repositories, configuration, stored credentials and the ability to publish or alter artifacts. Fastly and Wiz both describe the exploit as one unauthenticated request. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | self-hosted Artifactory (not JFrog Cloud) on a release below the first fixed build for its train | Run GET /artifactory/api/system/version against the instance (a valid user is enough, anonymous works if enabled) and read the version field. In the UI, the version shows in the footer or About page. Compare with the table; for example 7.146.25 and 7.161.19 are affected, 7.161.20 is fixed. JFrog Cloud was already fixed by JFrog; self-hosted means you run the service or container yourself. | | the join key setting is still the default, with no additionalJoinKeys value, and the Access join endpoint /access/api/v1/registry/join answers requests from the attacker's network without credentials | On the node, open system.yaml and look under shared, security for an additionalJoinKeys entry; for containers and Helm, check the environment variable JF_SHARED_SECURITY_ADDITIONALJOINKEYS. JFrog's workaround is to add that setting, and Fastly says the empty configuration value on a default install is what lets the empty key in. Then, from an untrusted network position, confirm that /access/api/v1/registry/join answers through your reverse proxy or load balancer. Wiz and Fastly saw exploitation as a POST to that path; no credentials are needed. | ## Affected and fixed versions | Release train | Affected (JFrog) | First fixed | | --- | --- | --- | | Before 7.111.4 (7.111.0 to 7.111.3, and older trains such as 7.104) | Not listed by JFrog or NVD as affected | No fixed build is named for this CVE | | 7.111.x | 7.111.4 to 7.111.20 | 7.111.21 | | 7.117.x | 7.117.0 to 7.117.27 | 7.117.28 | | 7.125.x | 7.125.0 to 7.125.19 | 7.125.20 | | 7.133.x | 7.133.0 to 7.133.28 | 7.133.29 | | 7.146.x | 7.146.0 to 7.146.36 (JFrog's table); 7.146.37 is in neither JFrog column, and NVD treats everything below 7.146.38 as affected | 7.146.38 | | 7.161.x | 7.161.0 to 7.161.19 | 7.161.20 | | Trains not listed above (for example 7.141) | Not listed by JFrog or NVD as affected | No fixed build is named for this CVE | JFrog's affected column shows ranges such as 7.111.4 > 7.111.21 whose end equals the fixed build on the 7.111 train; its patched column and NVD both give 7.111.21 as the first fixed build, so 7.111.21 is not affected. For 7.146 JFrog's affected range ends at 7.146.36 and its patched version is 7.146.38; 7.146.37 appears as neither affected nor patched, and NVD treats it as affected, so upgrade to 7.146.38. JFrog and NVD name affected builds only in the six ranges above. They name no affected build below 7.111.4 and no train other than 7.111, 7.117, 7.125, 7.133, 7.146 and 7.161, and JFrog publishes no statement that those builds are safe. If you run one, plan the upgrade to the nearest fixed train and apply the join key workaround until then. The user-agent strings in Fastly's data (including JFrog-Access/7.141.17) are values the attacker sends, so they say nothing about which versions are vulnerable. ## If the fix cannot go in yet - Upgrade to the fixed build for your train (table above). This is the vendor's most reliable known remediation. Cost: a service restart and upgrade testing. - Workaround if you cannot upgrade quickly (JFrog): generate your own random hex join key with openssl rand -hex 16, add it as additionalJoinKeys under shared, security in system.yaml (or set JF_SHARED_SECURITY_ADDITIONALJOINKEYS in containers), keep your existing joinKey entry, then restart the Access service or the whole JPD. Your existing join key keeps working, so cluster operation is unaffected. Cost: a restart, and the new key must be kept as secret as the join key. - Patching does not revoke tokens already issued. Rotate the platform join key and revoke access tokens issued since 28 Aug 2026 (Fastly). Cost: every CI job, deploy pipeline and integration that holds a revoked token gets HTTP 401 until it is given a new one, so list those clients first and schedule the change; the join key rotation also needs the Access service restart. - Reduce exposure: restrict network access to Artifactory and the Access paths to trusted systems (Wiz). Fastly Next-Gen WAF customers can enable the vendor's CVE-2026-82329 virtual patch rule. Cost: a WAF rule is a compensating control only and does not remove the flaw. ## What to look for - Any POST to /access/api/v1/registry/join (also /join/router and /join/router?override=true) that returns HTTP 201 and carries a token whose key ID (kid) is e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855 (SHA-256 of an empty string) is high-confidence evidence of compromise (Fastly). The kid is in the header of the JWT in the POST body, so it shows only where request bodies are captured: a WAF, a reverse proxy with body logging, or a packet capture. Artifactory's request*.log files in its log directory record the method, path and status, so there look for POST requests to the join paths that returned 201 from addresses that are not your cluster nodes. A normal cluster peer signs with a real join key and shows a different kid. Wiz saw 200 or 201 with an admin-scoped token in the response. User-agent indicators are easy to change. Cluster nodes legitimately call the join endpoint, so compare each source address with your node list before treating a 201 without that kid as an attack. - Fastly user-agent strings: jfrt/0/7.161.19, JFrog-Access/7.141.17, JFrogArtifactory/7.146.25. Sample high-volume sources (Fastly): 213.136.72[.]87, 89.117.49[.]147, 45.33.75[.]219, 23.92.20[.]11, 173.255.226[.]148, 195.154.157[.]234, 62.210.87[.]212, 45.79.190[.]169, 163.172.132[.]245. - Follow-on calls to correlate (Wiz): GET /api/system/configuration, PUT /api/security/users/ with an admin flag in customData, token minting at /access/api/v1/tokens, reads of /access/api/v1/system/security/join_key, and enumeration of users, repositories and tokens. Attackers also attached their own SSH keys to the users they created. - Account names seen (Wiz): Nxploited_ plus 3 characters, labadmin_ plus hex, 0xTerror, svc_ plus 8 characters, and look-alike names such as jfrog-distribution, jfrog-insight and repo-service. - Files: Groovy plugins in $JFROG_HOME/var/etc/artifactory/plugins/ (rce.groovy, cmd.groovy, jwtToken.groovy, uploadFile.groovy, JFrogImage.groovy, metrics.groovy, httpSession.groovy), exposed at /artifactory/api/plugins/execute/. A broken plugin API, or requests to registry/join, api/plugins or systemUpload rejected with HTTP 400, can mean a filter called traceFilter, held in memory, is installed (Wiz). Upgrading leaves these plugins and filters in place. - Attackers may truncate request*.log files, so empty or missing request logs are themselves an indicator (Wiz). If Access signing keys (private.key) were read, reset the token certificate; a patch does not revoke forged tokens. ## Timeline | Date | Event | Source | | --- | --- | --- | | 28 Aug 2026 | JFrog publishes the advisory and NVD publishes the record. Fastly sees no exploitation on disclosure day, and only single-digit probes on 29 and 30 Aug 2026. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 31 Aug 2026 | About 75,000 requests, nearly 98% from offensive-security services (Fastly). | [source](https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity) | | 1 Sep 2026 | A public exploit appears; Fastly sees about 171,000 attempts and mass scanning. Wiz's observed successful exploitation runs 1 to 8 Sep 2026. | [source](https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity) | | 2 Sep 2026 | CISA adds the CVE to KEV; Fastly records its peak of about 406,000 attempts. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-82329) | | 3 Sep 2026 | Fastly publishes its analysis and WAF virtual patch; NVD record modified. | [source](https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity) | | 5 Sep 2026 | CISA KEV due date, three days after the add. The KEV required action is to apply mitigations in accordance with vendor instructions, follow CISA's Forensics Triage Requirements and discontinue use of the product if mitigations are unavailable. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-82329) | | 10 Sep 2026 | Wiz publishes its report on this CVE and CVE-2026-42016 and CVE-2026-42018. It corrects the HTTP method for this CVE on 11 Sep 2026 and adds the malicious Groovy plugin section and more indicators on 17 Sep 2026. | [source](https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201) | Exploited 4 days after disclosure: disclosed 28 Aug 2026, first exploitation reported 1 Sep 2026. ## Exploitation reported so far Fastly Security Research reported on 3 Sep 2026 that probing began on 31 Aug 2026, mostly from offensive-security vendors, and that mass scanning followed on 1 Sep 2026 when a public exploit appeared. Individual IPs sprayed thousands of hosts, and sources grew to nearly 1,400 addresses in a day. Fastly counted about 406,000 attempts on 2 Sep 2026. Wiz Research confirmed successful exploitation between 1 and 8 Sep 2026 by several actors, each doing different things: configuration theft, new admin accounts, long-lived tokens, join key theft, attached SSH keys and Groovy plugins. CISA added it to KEV on 2 Sep 2026 with a due date of 5 Sep 2026, flagged for forensic triage; ransomware use is listed as unknown. ## Why agent investigation settles it The version and the default join-key configuration show whether the instance could be hit, but the fix does not evict an attacker who already minted tokens, created users or planted plugins. For an instance that was reachable while vulnerable the useful question is whether it was compromised, which needs log and file investigation against the join endpoint and plugins directory. ## Related pages - [CVE-2026-80521: Linux kernel](/ledger/cve-2026-80521/) - [CVE-2026-84869: ConnectWise ScreenConnect](/ledger/cve-2026-84869/) - [CVE-2026-75604: Vercel Next.js](/ledger/cve-2026-75604/) ## Sources - CVE-2026-82329: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-82329 (read 19 Nov 2026) - CVE-2026-82329: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-82329 (read 19 Nov 2026) - CVE-2026-82329: JFrog security advisories (vendor advisory, CVE-2026-82329 section and workaround): https://docs.jfrog.com/releases/docs/jfrog-security-advisories (read 19 Nov 2026) - CVE-2026-82329: Fastly Security Research report, 3 Sep 2026: https://www.fastly.com/blog/cve-2026-82329-jfrog-artifactory-authentication-bypass-exploitation-activity (read 19 Nov 2026) - CVE-2026-82329: Wiz Research report on CVE-2026-42016, CVE-2026-42018 and CVE-2026-82329 (the canonical URL as Wiz publishes it), published 10 Sep 2026, corrected 11 Sep 2026, extended 17 Sep 2026: https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201 (read 19 Nov 2026) - CVE-2026-82329: JFrog documentation: Get Artifactory Version API: https://docs.jfrog.com/administration/reference/getartifactoryversion (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-82329/ --- # CVE-2026-80521: Linux kernel, is it exploitable where you run it Ledger entry, disclosed 26 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-80521 affects Linux kernel and was disclosed 26 Aug 2026 with a CVSS score of 7.8. Exposure needs these conditions: the running kernel has the new UNIX socket garbage collector and lacks the fix; an attacker already runs code on the host as an unprivileged local user. Attack-path analysis settles the question for this bug. ## What the flaw is The kernel passes open sockets between processes over UNIX domain sockets. A garbage collector frees groups of such sockets that nothing can reach any more. Since Linux 6.10 it tracks the groups as strongly connected components. Kyle Zeng reported that a socket sent to itself while its group is being closed can leave the collector in a half-finished state. One group is judged dead but not freed, because a message that holds a reference is not yet queued. The next collection run then walks to a freed entry through its scc_entry list link. The result is a use-after-free in the kernel. Red Hat rates the flaw Important and says an authenticated local user can use the race to corrupt kernel memory, with a possible local privilege escalation. The fix adds one line that unlinks scc_entry in unix_del_edge() before the entry is freed. CISA's KEV catalog and the vendor trackers list no exploitation of this flaw. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the running kernel has the new UNIX socket garbage collector and lacks the fix | Run `uname -r` and compare with the versions table. The collector is in mainline 6.10 or later, in 6.6.93 or later and 6.1.141 or later in those lines, and in Ubuntu's 6.8 kernels. A distribution kernel can carry the fix under an older version number, so on RPM systems run `rpm -q --changelog kernel \| grep CVE-2026-80521`, and on Debian and Ubuntu run `apt changelog linux-image-$(uname -r) \| grep CVE-2026-80521`, or use the Debian and Ubuntu trackers. | | an attacker already runs code on the host as an unprivileged local user | Unprivileged users can create UNIX domain sockets by default (CVSS AV:L, PR:L). List who can log in or execute code: `getent passwd`, `last`, service accounts, CI runners and container workloads. Without a local foothold this CVE is not reachable. | ## Affected and fixed versions | Product or line | Affected | First fixed | | --- | --- | --- | | Linux mainline | 6.10 up to and including 7.1 (rewrite landed in 6.10, 14 Jul 2024; 7.1 stable is fixed from 7.1.10, see below) | 7.2 (16 Aug 2026), commit 594d905195024b228c962627ae5ae7c17bd582a4 | | Linux 7.1 stable | All earlier 7.1.x | 7.1.10 (23 Aug 2026) | | Linux 6.18 stable | All earlier 6.18.x | 6.18.53 (21 Sep 2026) | | Linux 6.12 stable | All earlier 6.12.x | 6.12.111 (21 Sep 2026) | | Linux 6.6 stable | 6.6.93 and later before the fix | 6.6.158 (3 Oct 2026) | | Linux 6.1 stable | 6.1.141 and later before the fix | 6.1.189 (3 Oct 2026) | | Linux 6.19 and 7.0 stable | 6.19.x and 7.0.x (both branches ended at 6.19.14 and 7.0.14) | No kernel.org fixed release. Ubuntu fixed its own 7.0 kernel (resolute 7.0.0-42.42) | | Linux 5.15 stable | Not affected (no new garbage collector) | Not applicable | | RHEL 10 | All earlier kernels | 10.2: kernel-6.12.0-211.65.1.el10_2 (RHSA-2026:81406, 12 Oct 2026). 10.3: kernel-6.12.0-264.14.1.el10_3 (RHSA-2026:98608, 11 Nov 2026). 10.0 EUS: kernel-6.12.0-55.109.1.el10_0 (RHSA-2026:77634, 8 Oct 2026) | | RHEL 6, 7, 8 and 9 | Not affected (Red Hat) | Not applicable | | Debian | bookworm, trixie and unstable before the fix | bookworm 6.1.190-1 (DLA-4841-1, upstream 6.1.190 includes the 6.1.189 fix), linux-6.12 on bookworm 6.12.111-1~deb12u1 (DLA-4817-1), trixie 6.12.111-1 (DSA-6528-1), unstable 7.1.10-1. Bullseye is not affected | | Ubuntu | noble 6.8 (Ubuntu backported the garbage collector rewrite in 6.8.0-98.98), resolute 7.0 and jammy HWE 6.8 before the fix | noble 6.8.0-149.149, jammy HWE 6.8.0-149.149~22.04.1, resolute 7.0.0-42.42. Focal, bionic, xenial and jammy 5.15 are not affected | NVD has not analyzed this record and lists no version ranges. The ranges above come from the kernel.org change logs and the vendor trackers. The 6.1.141 and 6.6.93 change logs (4 Jun 2025) both list "af_unix: Replace garbage collection algorithm", which is why those releases are the start of the affected range in their lines. Ubuntu 6.8 kernels are affected although 6.8 is below the mainline start of 6.10: the Ubuntu noble changelog for 6.8.0-98.98 lists the backport of the af_unix garbage collection rewrite, and 6.8.0-149.149 lists this fix. Ubuntu's upstream line says the fix is in 7.2 and 7.0.y, but the kernel.org change logs for 7.0.1 through 7.0.14, the last 7.0 release, do not contain it; Ubuntu's 7.0.0-42.42 kernel carries the fix as its own backport. ## If the fix cannot go in yet - Install a kernel with the fix and reboot. Cost: a reboot. This is the only remedy the vendors give. - Red Hat states that a mitigation is either not available or does not meet its criteria for ease of use, reach and stability. Neither Red Hat nor the kernel project publishes a workaround, and CISA has issued no guidance because the flaw is not in KEV. - Limiting local code execution is the only exposure-reducing step: restrict shell and CI runner access, and keep untrusted workloads off the host. User-namespace limits and seccomp profiles that still allow AF_UNIX do not remove the flaw, because AF_UNIX sockets need no privilege and no namespace. A seccomp filter that denies `socket(AF_UNIX, ...)` would block the path, but it breaks syslog, D-Bus, systemd clients, X11 and most container runtimes, and neither Red Hat nor the kernel project recommends it. ## What to look for - No indicators of compromise have been published by Red Hat, Debian, Ubuntu or the kernel project, and the flaw is not in the CISA KEV catalog (feed version 2026.11.18). A successful trigger would show as a kernel oops or panic in `dmesg` or `journalctl -k` that names unix_gc or the scc_entry list, and a host that reboots without a planned change. ## Timeline | Date | Event | Source | | --- | --- | --- | | 14 Jul 2024 | Linux 6.10 is released with the new af_unix garbage collector (commit 4090fa373f0e) that the fix later cites as the origin of the bug. | [source](https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.10) | | 4 Jun 2025 | Stable releases 6.1.141 and 6.6.93 backport the garbage collector. | [source](https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.6.93) | | 4 Aug 2026 | Kuniyuki Iwashima writes the fix after a report by Kyle Zeng, who reviews it. | [source](https://api.github.com/repos/torvalds/linux/commits/594d905195024b228c962627ae5ae7c17bd582a4) | | 6 Aug 2026 | The fix is committed to the netdev tree by Jakub Kicinski. | [source](https://api.github.com/repos/torvalds/linux/commits/594d905195024b228c962627ae5ae7c17bd582a4) | | 16 Aug 2026 | Linux 7.2 is released with the fix. | [source](https://cdn.kernel.org/pub/linux/kernel/v7.x/ChangeLog-7.2) | | 23 Aug 2026 | Stable 7.1.10 ships the fix. | [source](https://cdn.kernel.org/pub/linux/kernel/v7.x/ChangeLog-7.1.10) | | 26 Aug 2026 | The kernel CNA publishes CVE-2026-80521. NVD publishes the record (CVSS 7.8 from the CNA) and Red Hat sets its public date. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-80521) | | 21 Sep 2026 | Stable 6.12.111 and 6.18.53 ship the fix. | [source](https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.12.111) | | 3 Oct 2026 | Stable 6.1.189 and 6.6.158 ship the fix. | [source](https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.1.189) | | 12 Oct 2026 | Red Hat ships RHSA-2026:81406 for RHEL 10.2; the 10.3 errata RHSA-2026:98608 follows on 11 Nov 2026. | [source](https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-80521.json) | No public report dates the start of exploitation. ## Why attack-path analysis settles it The version and distribution tables settle whether the kernel has the new collector. The open question is whether anyone can run code on the host as a local user, which an attack-path view answers from who and what can execute there. No exploitation is known, so a known-exploited feed has nothing to say yet. ## Related pages - [CVE-2026-75604: Vercel Next.js](/ledger/cve-2026-75604/) - [CVE-2026-82329: JFrog Artifactory](/ledger/cve-2026-82329/) - [CVE-2026-19490: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-19490/) ## Sources - CVE-2026-80521: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-80521 (read 19 Nov 2026) - CVE-2026-80521: CISA KEV feed (catalog version 2026.11.18), no entry for this CVE: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - CVE-2026-80521: vendor advisory data (Red Hat Security Data API: errata, package versions, statement): https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-80521.json (read 19 Nov 2026) - CVE-2026-80521: upstream fix commit (GitHub mirror of the Linux kernel); the five stable commits in the NVD references were read the same way: https://api.github.com/repos/torvalds/linux/commits/594d905195024b228c962627ae5ae7c17bd582a4 (read 19 Nov 2026) - CVE-2026-80521: kernel.org stable change log 6.1.189 (also read for 6.6.158, 6.12.111, 6.18.53, 7.1.10, 7.2, 6.1.141, 6.6.93 and 6.10): https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.1.189 (read 19 Nov 2026) - CVE-2026-80521: Debian security tracker: https://security-tracker.debian.org/tracker/CVE-2026-80521 (read 19 Nov 2026) - CVE-2026-80521: Ubuntu CVE tracker data: https://ubuntu.com/security/cves/CVE-2026-80521.json (read 19 Nov 2026) - CVE-2026-80521: GitHub advisory database entry GHSA-g8mr-r59r-2xmq: https://api.github.com/advisories?cve_id=CVE-2026-80521 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-80521/ --- # CVE-2026-75604: Vercel Next.js, is it exploitable where you run it Ledger entry, disclosed 25 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-75604 affects Vercel Next.js and was disclosed 25 Aug 2026 with a CVSS score of 9.0. Exposure needs these conditions: next is 13.4.0 to before 15.5.24, or 16.0.0 to before 16.3.3; a Windows server answers unauthenticated requests for a Pages Router or non-Cache-Components App Router app. It was not in the CISA Known Exploited Vulnerabilities catalog on the read date. ## What the flaw is Next.js builds file paths for its incremental cache from the route segments of a request. On servers that use a Windows filesystem, it does not consistently escape backslashes in those segments. A request that carries encoded Windows path separators can therefore step outside the cache directory and read files that should stay private. The files at risk include private build data. NVD names the server-reference-manifest encryption key. The vendor rates the result as unauthenticated remote code execution: with that key an attacker can act against the application's server actions. The CVSS vector sets attack complexity to High, and the scope as changed. The flaw sits in the shared path-escaping helper (escape-path-delimiters.ts) and the file-system cache (file-system-cache.ts). The cache joins the request's route segments onto three roots: .next/server/pages for the Pages Router, .next/server/app for App Router pages and routes, and .next/cache/fetch-cache for fetch data. A segment that carries an encoded Windows separator can climb out of those roots, so the files reachable are files under the application's build output, for example server-reference-manifest.json with its encryption key. The fix escapes backslashes in segments and rejects any resolved cache path outside its root (error text "Invalid file path"). It applies to the Pages Router and to the App Router when Cache Components is not in use. Linux and macOS hosts are not named as affected. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | next is 13.4.0 to before 15.5.24, or 16.0.0 to before 16.3.3 | In the project folder run `npm ls next` (or `pnpm why next`, `yarn why next`). For a deployed bundle run `node -p "require('next/package.json').version"` from the app directory on the server. Compare with the versions table; 13.x and 14.x have no patch, so move to 15.5.24 or later. Check every deployed copy, not only the repository, because a lockfile can differ from what is running. | | a Windows server answers unauthenticated requests for a Pages Router or non-Cache-Components App Router app | The flaw affects an app that uses the Pages Router or the App Router without Cache Components. Look for a `pages` or `src/pages` folder (Pages Router) and an `app` or `src/app` folder (App Router) in the project. For Cache Components, open next.config.js or next.config.ts and look for the `cacheComponents` setting. An App Router app with that setting on is outside the affected configuration the vendor lists; an app without it is inside. Windows: On the host run `node -p process.platform`: the value `win32` means Windows. In PowerShell `(Get-CimInstance Win32_OperatingSystem).Caption` names the edition. For a container run `docker inspect --format '{{.Os}}' `: `windows` means a Windows container. A Linux container on a Windows laptop uses a Linux filesystem and is outside the condition the vendor describes. Reach (CVSS: network, no privileges, no user interaction): find which addresses and ports the Next.js server listens on (`netstat -ano \| findstr LISTENING` on Windows) and whether the firewall, load balancer or reverse proxy publishes them. An internal-only service still counts for anyone who can reach it. | | for code execution, the app uses server actions, so the leaked server-reference-manifest encryption key is usable (NVD: disclosure of that key can enable remote code execution) | Search the source for the "use server" directive. The fix's own test notes that a server action makes next build emit server-reference-manifest.json, the file that holds the encryption key. An application with no server actions emits no such manifest, so there is no key to leak, but other private build JSON under the .next output remains readable. The vendor still rates the CVE as remote code execution. | ## Affected and fixed versions | Release line | Affected versions | First fixed version | | --- | --- | --- | | Next.js 13.4 and later 13.x, and 14.x | All, from 13.4.0 | None for these lines: upgrade to 15.5.24 or later | | Next.js 15.x | 15.0.0 up to but excluding 15.5.24 | 15.5.24 | | Next.js 16.x | 16.0.0 up to but excluding 16.3.3 | 16.3.3 | The advisory lists two patched releases, 15.5.24 and 16.3.3, and no patched 13.x or 14.x release. Users of 13.4.x and 14.x have no backport and must jump to 15.5.24 or later, which is a major-version upgrade. Versions before 13.4.0 are not listed as affected. ## If the fix cannot go in yet - Upgrade to Next.js 15.5.24 or 16.3.3, then rebuild and redeploy; 13.4.x and 14.x apps must move to 15.5.24 or later. This is the only fix. Cost: a rebuild and a redeploy, plus regression testing for a version bump. - The vendor states there is no known workaround for affected Windows-hosted applications and says to upgrade immediately if the server is hosted on Windows. - Until the upgrade ships, reduce exposure by keeping affected Windows-hosted instances off the internet (allow only trusted networks at the firewall or reverse proxy). Cost: users outside those networks lose access. This step is our guidance, not the vendor's. ## What to look for - No indicators of compromise or log strings were published by the vendor, NVD or CISA, and CISA's SSVC data in NVD records exploitation as none. The patch shows what the abuse looks like in a web or reverse proxy access log (derived from the fix, not vendor guidance): requests to page routes or /_next/data/ paths that contain an encoded backslash (%5c or %5C) in a route segment. A search that matches that shape is the regular expression `/_next/data/[^ ]*%5[cC]` or, for page routes, `GET /[^ ]*%5[cC]`. Ordinary routes carry none, so any hit on a Windows-hosted instance is worth review; the status code alone does not show whether a file was read, so check what the response returned. - If an affected Windows-hosted instance was reachable from untrusted networks, treat build secrets, including the server-reference-manifest encryption key, as exposed and rotate them by rebuilding after the upgrade. This is our guidance, not the vendor's. ## Timeline | Date | Event | Source | | --- | --- | --- | | 25 Aug 2026 | Next.js 15.5.24 and 16.3.3 are released (published 16:16 and 16:17 UTC) with the fix; the release notes link this advisory as critical remote code execution on Windows-hosted servers. This is the vendor's disclosure date. | [source](https://github.com/vercel/next.js/releases/tag/v16.3.3) | | 1 Sep 2026 | NVD publishes the CVE record (source: GitHub). CISA's SSVC entry, added 3 Sep, records exploitation as none. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604) | | 8 Sep 2026 | GitHub publishes the reviewed record of advisory GHSA-p293-qw3h-jr36, severity critical, credits evolutionstorm and B0RI as reporters. | [source](https://api.github.com/advisories/GHSA-p293-qw3h-jr36) | No public report dates the start of exploitation. ## Why static reachability settles it Exposure needs facts that can be read from the deployment: an affected version, a Windows filesystem, a router setup without Cache Components and, for code execution, server actions that make the encryption key exist. A static inventory answers all four without exploitation data. The vendor gives no workaround and the fix is a version change, so the checks table settles it. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-19490: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-19490/) - [CVE-2026-80521: Linux kernel](/ledger/cve-2026-80521/) - [CVE-2026-64849: MLflow Tracking Server](/ledger/cve-2026-64849/) ## Sources - CVE-2026-75604: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604 (read 19 Nov 2026) - CVE-2026-75604: vendor advisory (Next.js GHSA-p293-qw3h-jr36, via the GitHub advisory API): https://github.com/vercel/next.js/security/advisories/GHSA-p293-qw3h-jr36 (read 19 Nov 2026) - CVE-2026-75604: GitHub advisory API record for the Next.js advisory: https://api.github.com/advisories/GHSA-p293-qw3h-jr36 (read 19 Nov 2026) - CVE-2026-75604: Next.js 15.5.24 release notes: https://github.com/vercel/next.js/releases/tag/v15.5.24 (read 19 Nov 2026) - CVE-2026-75604: Next.js 16.3.3 release notes: https://github.com/vercel/next.js/releases/tag/v16.3.3 (read 19 Nov 2026) - CVE-2026-75604: vendor fix commit for the 16.3 line: https://github.com/vercel/next.js/commit/968b9fcb26bdeb8e0a861a9df05361474666d51b (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-75604/ --- # CVE-2026-19490: Citrix NetScaler ADC and Gateway, is it exploitable where you run it Ledger entry, disclosed 19 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-19490 affects Citrix NetScaler ADC and Gateway and was disclosed 19 Aug 2026 with a CVSS score of 9.3. Exposure needs these conditions: the NetScaler ADC or Gateway build is before 14.1-73.32, 13.1-63.21, 14.1-73.32 FIPS or 13.1-37.277; it is configured as a Gateway or AAA virtual server, and also has a SAML action on newer builds. ## What the flaw is CVE-2026-19490 is an authentication bypass using an alternate path or channel (CWE-288) in NetScaler ADC and NetScaler Gateway. An unauthenticated attacker who can reach a Gateway or AAA virtual server over the network may bypass authentication. Citrix rates it 9.3 on CVSS 4.0 (network, low complexity, no privileges, no user interaction, high impact on confidentiality, integrity and availability of the appliance). Whether a given appliance is exposed depends on its build and its configuration. The flaw applies only to appliances set up as a Gateway (SSL VPN, ICA Proxy, CVPN, RDP Proxy) or as an AAA virtual server. On newer builds Citrix narrows this further to appliances that also use a SAML action. Citrix says the bulletin covers customer-managed appliances only: Citrix-managed cloud services and Citrix-managed Adaptive Authentication are updated by Cloud Software Group. The same bulletin, CTX696939, also covers CVE-2026-19489, a separate memory overflow that needs SIP ALG enabled on a Large Scale NAT group. The two share the same fixed builds. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the NetScaler ADC or Gateway build is before 14.1-73.32, 13.1-63.21, 14.1-73.32 FIPS or 13.1-37.277 | On the appliance CLI run `show ns version` and compare the build with the versions table (14.1 before 14.1-73.32, 13.1 before 13.1-63.21, 14.1 FIPS before 14.1-73.32 FIPS, 13.1 FIPS and NDcPP before 13.1-37.277). Check every node of an HA pair or cluster. | | it is configured as a Gateway or AAA virtual server, and also has a SAML action on newer builds | Citrix says to inspect the configuration for these strings. Gateway virtual server (SSL VPN, ICA Proxy, CVPN or RDP Proxy): `show ns runningConfig \| grep "add vpn vserver"`. AAA virtual server: `show ns runningConfig \| grep "add authentication vserver"`. SAML action: `show ns runningConfig \| grep "add authentication samlAction"`. Any hit on the first two means a Gateway or AAA virtual server is present. Which builds also need the SAML hit: on 14.1-43.56 or later, 14.1-66.68-FIPS or later and 13.1-61.28 or later (non-FIPS), Citrix lists the flaw as applicable only with a SAML action, so no samlAction hit means those builds are not exposed. On 14.1-43.55 or earlier, 13.1-61.27 or earlier and 13.1-FIPS builds, a Gateway or AAA virtual server alone is enough. The bulletin states the 13.1-FIPS rule without a SAML requirement. It lists no rule for 13.1-NDcPP, so assume a Gateway or AAA virtual server alone is enough on that build. | | the virtual server is reachable by the attacker, who needs no credentials | List the virtual server addresses and ports with `show vpn vserver` and `show authentication vserver`, then check which are published to the internet or to untrusted networks in your firewall rules. | ## Affected and fixed versions | Product line | Affected builds | First fixed build | | --- | --- | --- | | NetScaler ADC and Gateway 14.1 | Before 14.1-73.32 | 14.1-73.32 | | NetScaler ADC and Gateway 13.1 | Before 13.1-63.21 | 13.1-63.21 | | NetScaler ADC 14.1-FIPS | Before 14.1-73.32 FIPS | 14.1-73.32 FIPS | | NetScaler ADC 13.1-FIPS and 13.1-NDcPP | Before 13.1-37.277 | 13.1-37.277 | Each fixed build and all later releases of the same train are fixed. NVD words the range as 'through 73.32' and 'through 63.21', but the Citrix bulletin says 'before' those builds and names them as the fixes. Secure Private Access Hybrid deployments need their NetScaler instances moved to these builds too. ## If the fix cannot go in yet - No workaround: Citrix lists 'None' under workarounds and mitigating factors. The fix is the first fixed build for your train (see the versions table). Plan for the restart an ADC upgrade normally needs: the bulletin states no downtime, no upgrade order for HA pairs or clusters and no step to end existing sessions. - Not from Citrix: until the upgrade is in, restrict which source addresses can reach the Gateway and AAA virtual servers, for example with firewall rules or ACLs that allow only known partner or office ranges. Cost: remote users outside those ranges lose VPN or ICA access, so this suits appliances with a known user population and not a public VPN. It narrows exposure and does not remove the flaw, because the vendor lists no workaround. - CISA's KEV entry requires applying the vendor's mitigations in line with BOD 26-04 and its forensics triage requirements, or discontinuing use of the product if mitigations are unavailable. The KEV entry marks forensic triage as Yes for this CVE. ## What to look for - Citrix and CISA publish no log strings, session artifacts or indicator addresses for this CVE. The bypass skips authentication, so a successful attack leaves no failed-login trail to search for. Treat an unpatched Gateway or AAA appliance that was reachable from 19 Aug 2026 as a candidate for triage. - Collect this evidence from each appliance before you upgrade, because the upgrade restarts it: /var/log/ns.log and the rotated ns.log.* files for Gateway logins (the SSLVPN LOGIN lines, which carry SessionId, User and Client_ip) and AAA virtual server login lines that have no matching authentication event from the identity provider (Citrix names no log message for this CVE, so these are NetScaler's general session records); the list of live AAA and Gateway sessions with `show aaa session` and the matching connection lists, compared with your user directory; and the performance-counter data that `nsconmsg` reads, saved with a configuration backup and a copy of `show ns runningConfig`. - Follow CISA's forensic triage steps for the rest: the KEV entry requires them under [BOD 26-04](https://www.cisa.gov/news-events/directives/bod-26-04-implementation-guidance-prioritizing-security-updates-based-risk), with target times that run from the KEV addition on 9 Sep 2026. ## Timeline | Date | Event | Source | | --- | --- | --- | | 19 Aug 2026 | Citrix publishes bulletin CTX696939 (initial publication, the only change-log entry) covering CVE-2026-19489 and CVE-2026-19490, with fixed builds. NVD publishes the CVE record the same day. The bulletin credits Samarth Vashisht of the JPMorgan Chase pen-test team. | [source](https://support.citrix.com/external/article/CTX696939/netscaler-adc-and-netscaler-gateway-secu.html) | | 9 Sep 2026 | CISA adds the CVE to the KEV catalog with a due date of 12 Sep 2026 and forensic triage marked yes. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-19490) | | 26 Oct 2026 | NVD last modifies the record (status Analyzed). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-19490) | The earliest dated mark of exploitation is CISA's, on 9 Sep 2026, 21 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far The KEV entry describes the flaw as an authentication bypass on appliances configured as an AAA virtual server or as a Gateway, and lists ransomware campaign use as Unknown. No report names the date of first attack or the attacker. ## Why static reachability settles it Exposure depends on the build and on whether the appliance runs a Gateway or AAA virtual server, plus a SAML action on newer builds. All of that is readable from the appliance with `show ns version` and a grep of the running configuration, so a configuration read settles it faster than a CVSS score or a feed. KEV confirms exploitation, which makes the check urgent but does not say whether your appliance is exposed. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-64849: MLflow Tracking Server](/ledger/cve-2026-64849/) - [CVE-2026-75604: Vercel Next.js](/ledger/cve-2026-75604/) - [CVE-2026-42018: JFrog Artifactory](/ledger/cve-2026-42018/) ## Sources - CVE-2026-19490: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-19490 (read 19 Nov 2026) - CVE-2026-19490: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-19490 (read 19 Nov 2026) - CVE-2026-19490: Citrix security bulletin CTX696939 (vendor advisory): https://support.citrix.com/external/article/CTX696939/netscaler-adc-and-netscaler-gateway-secu.html (read 19 Nov 2026) - CVE-2026-19490: CISA BOD 26-04 implementation guidance (forensic triage steps): https://www.cisa.gov/news-events/directives/bod-26-04-implementation-guidance-prioritizing-security-updates-based-risk (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-19490/ --- # CVE-2026-64849: MLflow Tracking Server, is it exploitable where you run it Ledger entry, disclosed 17 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-64849 affects MLflow Tracking Server and was disclosed 17 Aug 2026 with a CVSS score of 9.3. Exposure needs these conditions: MLflow is 3.3.0 to 3.14.x (3.15.x on IPv6 or NAT64) with a SQL store, the default; the webhooks API is open (no --app-name basic-auth) and the host can reach internal targets. ## What the flaw is The MLflow Tracking Server has a model-registry webhooks API. A webhook is a URL the server calls when a registry event happens, and a test endpoint (POST /api/2.0/mlflow/webhooks/{id}/test) calls the URL on demand and returns the response status and body to the caller. MLflow checks the webhook URL against a rule that rejects non-public addresses, but it checks only the URL as first written. Delivery then follows HTTP redirects and looks up the hostname again, without pinning the address it checked. An attacker points a webhook at a public HTTPS host they control. The host passes the check and then redirects to an internal address, or its DNS answer changes between the check and the call. The MLflow server fetches the internal target and the test endpoint hands the reply back. That reaches cloud instance metadata (which can hold temporary cloud credentials), internal admin services and internal ports. The advisory also describes a blind variant: redirect codes 307 and 308 keep the POST method and body, so an attacker can send POST requests to internal endpoints that act on POST. The default mlflow server runs with no authentication, so no login is needed. The only webhook authorization lives in an optional auth plugin that is not loaded by default. The CNA rates it 9.3 (CVSS 3.1 base score, vector CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:L/A:N: network, low complexity, no privileges, no user interaction, scope changed, high confidentiality impact, low integrity impact). ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | MLflow is 3.3.0 to 3.14.x (3.15.x on IPv6 or NAT64) with a SQL store, the default | Version: on the server host or inside its container run pip show mlflow, or mlflow --version. 3.3.0 to 3.9.x are exposed with no bypass needed, 3.10.0 to 3.14.x through the redirect and DNS bypass, 3.15.0 or later has the fix, and 3.16.0 or later is needed on IPv6 or NAT64 networks. Store: look at --backend-store-uri in the launch command (ps -ef \| grep 'mlflow server', or systemctl cat mlflow). No value means the default SQLite store, which qualifies. A file store only cannot hold webhooks (GitHub advisory notes). | | the webhooks API is open (no --app-name basic-auth) and the host can reach internal targets | Open API: look at the launch command or service unit (ps -ef \| grep 'mlflow server', or systemctl cat mlflow). The auth plugin is enabled only when the command includes --app-name basic-auth (MLflow authentication documentation); without that flag the API is open, because the only webhook authorization is in that optional plugin. To confirm, send the server any REST request without credentials from another host: if it is served instead of rejected, the API is open. Network reach: read the bind address in the launch command (--host), run ss -ltnp \| grep -i mlflow, then check the firewall, security group or reverse proxy rules for that port from an untrusted network. Outbound reach: from the server host run curl -s -m 3 -o /dev/null -w '%{http_code}\n' http://169.254.169.254/. On AWS, check whether IMDSv2 is required: aws ec2 describe-instances --instance-ids INSTANCE_ID --query 'Reservations[].Instances[].MetadataOptions.HttpTokens' (value required means IMDSv2 only). Review egress rules for the host. | ## Affected and fixed versions | Product line | Affected | First fixed | Notes | | --- | --- | --- | --- | | MLflow Tracking Server (pip package mlflow) | 3.3.0 up to but excluding 3.15.0 | 3.15.0 (released 31 Jul 2026) | Fix is pull request #24258 (connection-time check of the peer address). Webhooks exist from 3.3.0; the URL guard from 3.10.0 | | Follow-up hardening | 3.15.0 to 3.15.x | 3.16.0 | Pull request #25568 closes four IPv6-transition bypasses (NAT64, IPv4-mapped CGNAT) | MLflow 3.3.0 introduced model-registry webhooks (3.3.0 release notes). The URL guard that this CVE bypasses, pull request #20747 (validate webhook URL), was merged on 13 Feb 2026 and shipped in 3.10.0. Releases 3.3.0 to 3.9.x have webhooks but predate that guard, so they are affected without needing a bypass; 3.10.0 to 3.14.x carry the guard and are affected through the redirect and DNS-rebinding bypass. 3.15.0 checks the peer address at connection time. ## If the fix cannot go in yet - Upgrade to MLflow 3.15.0 or later, and to 3.16.0 or later on IPv6 or NAT64 networks. This is the fix. The advisory lists no workaround. - Exposure-reducing steps if you cannot upgrade yet. The advisory states that webhook authorization lives in the optional auth plugin, so restart the server with mlflow server --app-name basic-auth (admin password from MLFLOW_AUTH_ADMIN_PASSWORD) and keep it off untrusted networks. Cost: every client and script needs credentials, and users must sign in. - Require IMDSv2 (or the equivalent metadata protection) on the host and filter its outbound traffic, since the advisory names cloud instance metadata as the main target. Cost: software on the host that still uses IMDSv1 breaks, and webhooks to legitimate destinations need explicit allow rules. - CISA's KEV entry gives this required action: apply mitigations in accordance with vendor instructions, follow the BOD 26-04 guidance the entry cites, and discontinue use of the product if mitigations are unavailable. ## What to look for - No indicators of compromise have been published for this CVE. Search web server or reverse proxy access logs for webhook calls, for example grep -E 'POST /api/2.0/mlflow/webhooks(/[^ /]+/test)? ' access.log; a test call with a 200 status that you did not make is the pattern to investigate. List the registered webhooks with the API call GET /api/2.0/mlflow/webhooks (for example curl -s 'http://MLFLOW_HOST:5000/api/2.0/mlflow/webhooks?max_results=100') and read the url of each entry: any webhook that points at a host you do not recognize is a candidate. - Review egress and flow logs from the server host for connections to 169.254.169.254 or to private and loopback ranges that the server has no reason to contact. - If the host has a cloud role, review cloud audit logs for use of its temporary credentials from addresses outside your environment (general step). ## Timeline | Date | Event | Source | | --- | --- | --- | | 13 Feb 2026 | Pull request #20747 adds the webhook URL guard that this CVE later bypasses; it ships in MLflow 3.10.0. | [source](https://github.com/mlflow/mlflow/pull/20747) | | 12 Jun 2026 | freeman-bb reports the flaw privately through the GitHub advisory process. | [source](https://api.github.com/advisories?cve_id=CVE-2026-64849) | | 26 Jun 2026 | The same flaw is reported publicly in issue #24179 by an independent finder (the advisory credits the 12 Jun private report with discovery priority). | [source](https://github.com/mlflow/mlflow/issues/24179) | | 2 Jul 2026 | The fix, pull request #24258, is merged. | [source](https://github.com/mlflow/mlflow/pull/24258) | | 31 Jul 2026 | MLflow 3.15.0 is released with the fix. | [source](https://github.com/mlflow/mlflow/releases/tag/v3.15.0) | | 17 Aug 2026 | GitHub publishes the advisory GHSA-7gwp-5pfp-969j and NVD publishes CVE-2026-64849 (CVSS 3.1 base score 9.3). | [source](https://api.github.com/advisories?cve_id=CVE-2026-64849) | | 19 Aug 2026 | CISA adds the CVE to KEV with a due date of 2 Sep 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-64849) | | 4 Sep 2026 | MLflow 3.16.0 is released with pull request #25568 (merged 3 Sep 2026), which closes four IPv6-transition bypasses of the 3.15.0 fix. | [source](https://github.com/mlflow/mlflow/pull/25568) | The earliest dated mark of exploitation is CISA's, on 19 Aug 2026, 2 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far The KEV entry cites the fix pull request and the public issue, and lists ransomware campaign use as Unknown. Neither it nor the NVD record names the attackers, the targets or the first day of use, so attacks may have begun before the KEV listing. The CISA coordinator SSVC entry on the NVD record, stamped 20 Aug 2026, rates exploitation as active and the flaw as automatable. The entry is stamped one day after the KEV listing, so it adds no earlier date. ## Why runtime reachability settles it A version below 3.15.0 is not enough: exposure needs the webhooks API open without authentication, a network path to the server and outbound reach to something worth stealing. Those are facts about the running server and its host, which the checks table reads one by one. KEV marks the CVE as exploited, so a reachable open server is urgent. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-42018: JFrog Artifactory](/ledger/cve-2026-42018/) - [CVE-2026-19490: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-19490/) - [CVE-2026-65660: Microsoft SharePoint Server, SharePoint code injection](/ledger/cve-2026-65660/) ## Sources - CVE-2026-64849: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-64849 (read 19 Nov 2026) - CVE-2026-64849: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-64849 (read 19 Nov 2026) - CVE-2026-64849: vendor advisory (MLflow GitHub security advisory GHSA-7gwp-5pfp-969j, read through the GitHub advisory API record): https://github.com/mlflow/mlflow/security/advisories/GHSA-7gwp-5pfp-969j (read 19 Nov 2026) - CVE-2026-64849: GitHub advisory API record (affected range, dates, credits): https://api.github.com/advisories?cve_id=CVE-2026-64849 (read 19 Nov 2026) - CVE-2026-64849: vendor fix pull request: https://github.com/mlflow/mlflow/pull/24258 (read 19 Nov 2026) - CVE-2026-64849: public issue report: https://github.com/mlflow/mlflow/issues/24179 (read 19 Nov 2026) - CVE-2026-64849: vendor release notes for 3.15.0: https://github.com/mlflow/mlflow/releases/tag/v3.15.0 (read 19 Nov 2026) - CVE-2026-64849: MLflow authentication documentation: https://mlflow.org/docs/latest/self-hosting/security/basic-http-auth/ (read 19 Nov 2026) - CVE-2026-64849: vendor pull request that added the URL guard: https://github.com/mlflow/mlflow/pull/20747 (read 19 Nov 2026) - CVE-2026-64849: vendor release notes for 3.3.0: https://github.com/mlflow/mlflow/releases/tag/v3.3.0 (read 19 Nov 2026) - CVE-2026-64849: vendor follow-up pull request: https://github.com/mlflow/mlflow/pull/25568 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-64849/ --- # CVE-2026-42018: JFrog Artifactory, is it exploitable where you run it Ledger entry, disclosed 12 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-42018 affects JFrog Artifactory and was disclosed 12 Aug 2026 with a CVSS score of 7.5. Exposure needs these conditions: the instance is self-hosted Artifactory below the fixed build for its train (7.111.20, 7.117.27, 7.125.19, 7.133.28 or 7.146.8); the Access service is reachable by the attacker, whatever the anonymous access setting. In the CISA KEV catalog since 11 Sep 2026. ## What the flaw is Artifactory's Access service has an internal anonymous user. A request to one variant of the AWS token path made without credentials could return a token for that user, even on instances where anonymous access is switched off. JFrog classes it as improper authentication (CWE-287). The token only carries the rights of the anonymous user, which is why the CVSS base score is 7.5 (set by JFrog) with confidentiality impact only. It can still expose artifacts and repository data that the anonymous identity can read. Its main danger is as step one of a chain. A low-privileged token is exactly what CVE-2026-42016 needs to produce an administrator token. Together the two turn an unauthenticated request into admin control. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the instance is self-hosted Artifactory below the fixed build for its train (7.111.20, 7.117.27, 7.125.19, 7.133.28 or 7.146.8) | JFrog Cloud was fixed by JFrog, and 7.161.x is not listed as affected. Run GET /artifactory/api/system/version against the instance (a valid user is enough, anonymous works if enabled) and read the version field; the UI shows it in the footer or About page. A version below the build for its train in the versions table is affected. Self-hosted means you run the service or container yourself. | | the Access service is reachable by the attacker, whatever the anonymous access setting | Confirm /access/api/v1/ paths answer through your reverse proxy or load balancer from an untrusted network. Wiz saw the exploit use a POST to /access/api/v1/aws/token/ with a trailing slash. The flaw applies even with anonymous access disabled (Administration, Security, General), so turning it off does not clear you. | ## Affected and fixed versions | Release train | Affected | First fixed | | --- | --- | --- | | 7.111.x and earlier | Below 7.111.20 | 7.111.20 | | 7.117.x | 7.117.0 to below 7.117.27 | 7.117.27 | | 7.125.x | 7.125.0 to below 7.125.19 | 7.125.19 | | 7.133.x | 7.133.0 to below 7.133.28 | 7.133.28 | | 7.146.x | 7.146.0 to below 7.146.8 | 7.146.8 | | 7.161.x | Not listed as affected | Not applicable | JFrog's advisory writes the affected ranges as 7.117.0 to 7.117.27 and so on and lists that build as the patched version; NVD records each upper bound as 'before' that build, so the listed build is the first fixed one. The target builds are 7.111.21, 7.117.28, 7.125.20, 7.133.29, 7.146.38 and 7.161.20 and later: they are later than the builds above on every train, they are what Wiz gives as fixed for the CVE-2026-42016 and CVE-2026-42018 chain, and they are the builds JFrog gives for CVE-2026-82329 (advisory of 28 Aug 2026, in KEV), which has its own ledger entry. One upgrade to the target build closes all three. ## If the fix cannot go in yet - Upgrade to a fixed build for your train (versions table); the target builds in the note also close CVE-2026-42016 and CVE-2026-82329. JFrog lists no workaround for this CVE. Cost: a service restart and upgrade testing. - Restrict network access to the Access paths to trusted users and systems (Wiz). Cost: external consumers and remote users lose access unless they come through the allowed path. - After upgrading, hunt for persistence from earlier exploitation and reset the Access token certificate if you find any, because stolen signing keys keep working after a patch. Cost: a reset invalidates tokens and needs a planned window. ## What to look for - Highest-confidence log signature (Wiz): a 401 on the bare AWS token path (/access/api/v1/aws/token) followed by a 200 on a variant of it, from the same client, within a short window. Normal clients do not produce that pattern. - Chain follow-on: POST /access/api/v1/tokens returning 200, then requests whose actor is token:anonymous doing admin work such as PUT /api/security/users/ (Wiz). The sample grep for those requests is in the [CVE-2026-42016 entry](/ledger/cve-2026-42016/). - Actor IPs Wiz tied to CVE-2026-42018 exploitation: 186.247.79[.]240 (2 Sep 2026), 182.62.201[.]69 (4 Sep 2026), 121.123.141[.]155 (26 Aug 2026), and 149.102.229[.]150 (30 Aug 2026). Payload hosts, hashes and plugin names from the later stages of the chain are listed in the CVE-2026-42016 entry. - Attackers may truncate request*.log in the Artifactory log directory, so missing or empty logs are themselves a sign of tampering (Wiz). Also check for binaries in world-writable paths such as /dev/shm, /tmp and /var/tmp. ## Timeline | Date | Event | Source | | --- | --- | --- | | 27 Jul 2026 | JFrog publishes the companion bug CVE-2026-42016. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 12 Aug 2026 | JFrog publishes the CVE-2026-42018 advisory and NVD publishes the record. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 13 Aug 2026 | JFrog updates the advisory. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 15 Aug 2026 | Start of the window in which Wiz saw the CVE-2026-42018 and CVE-2026-42016 chain exploited (to 8 Sep 2026). | [source](https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201) | | 28 Aug 2026 | JFrog publishes CVE-2026-82329, whose fixed builds are later on every train and close this bug too. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 10 Sep 2026 | Wiz publishes its report. | [source](https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201) | | 11 Sep 2026 | CISA adds the CVE to KEV. Wiz corrects the fixed versions it gave for this CVE. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42018) | | 25 Sep 2026 | CISA KEV due date for federal agencies. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42018) | Exploited 3 days after disclosure: disclosed 12 Aug 2026, first exploitation reported 15 Aug 2026. ## Exploitation reported so far Wiz Research says several actors chained this bug with CVE-2026-42016 between 15 Aug and 8 Sep 2026, and knew of no earlier public report. The usual shape was an unauthenticated POST that returned an anonymous token, a second POST that turned it into an admin token, then an admin account created within minutes in some cases. Wiz also recovered a custom Rust backdoor and malicious Groovy plugins from compromised instances. Wiz measured that the share of organizations running Artifactory with a vulnerable instance fell only from 69% at publication (12 Aug 2026) to 62% four weeks later. Its figure for CVE-2026-42016 (59% six weeks after its 27 Jul 2026 publication) is a different CVE measured from a different date. ## Why attack-path analysis settles it On its own the bug yields only an anonymous-user token, so the version check shows it exists but not how much damage follows. Exposure becomes serious when the Access path is reachable and the same build is also open to CVE-2026-42016, which turns the token into an admin token. That is a question about the path an attacker can walk, so attack-path analysis fits. ## Related pages - [CVE-2026-65660: Microsoft SharePoint Server, SharePoint code injection](/ledger/cve-2026-65660/) - [CVE-2026-64849: MLflow Tracking Server](/ledger/cve-2026-64849/) - [CVE-2026-68820: Microsoft Windows AFD for WinSock](/ledger/cve-2026-68820/) ## Sources - CVE-2026-42018: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-42018 (read 19 Nov 2026) - CVE-2026-42018: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42018 (read 19 Nov 2026) - CVE-2026-42018: JFrog security advisories (vendor advisory, CVE-2026-42018 section): https://docs.jfrog.com/releases/docs/jfrog-security-advisories (read 19 Nov 2026) - CVE-2026-42018: Wiz Research report, 10 Sep 2026, updated 17 Sep 2026 (the URL as Wiz publishes it, which NVD lists too): https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201 (read 19 Nov 2026) - CVE-2026-42018: JFrog documentation: Get Artifactory Version API: https://docs.jfrog.com/administration/reference/getartifactoryversion (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-42018/ --- # CVE-2026-65660: Microsoft SharePoint Server, is it exploitable where you run it Ledger entry, disclosed 11 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-65660 affects Microsoft SharePoint Server and was disclosed 11 Aug 2026 with a CVSS score of 8.8. Exposure needs these conditions: a farm server is below the fixed SharePoint build in the versions table; the attacker has a low-privilege account or reaches an anonymous site. CISA added it to the Known Exploited Vulnerabilities catalog on 25 Sep 2026. ## What the flaw is CVE-2026-65660 is a code injection flaw (CWE-94) in on-premises SharePoint Server. Microsoft says an authenticated attacker with low-level access to an affected server can send a specially crafted request that runs code on the server. No user interaction is needed. The CVSS 3.1 base score is 8.8 (network, low complexity, low privileges, high impact on confidentiality, integrity and availability). Microsoft first published the CVE on 11 Aug 2026 as a spoofing flaw with a lower CVSS score. A revision on 27 Aug 2026 changed the title, impact and FAQs to remote code execution. Microsoft calls that revision informational, and the security update was available from 11 Aug 2026. Teams that triaged it as a medium spoofing issue in August should re-rate it. On its own the bug needs a logged-in account. Previdian reports that the attempts it saw paired it with a second, separate weakness that lets an unauthenticated request reach the same processing on sites that allow anonymous viewing. Previdian, citing Viettel Cyber Security research, says that second weakness was fixed on 9 Jun 2026 and has no CVE identifier in that write-up. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | a farm server is below the fixed SharePoint build in the versions table | On a SharePoint server, open the SharePoint Management Shell and run (Get-SPFarm).BuildVersion. Or open Central Administration > Upgrade and Migration > Check product and patch installation status. Compare with 16.0.5565.1001 (2016), 16.0.10417.20198 (2019) or 16.0.19725.20522 (Subscription Edition). Check every server, since a partial install leaves the farm exposed. Where Microsoft lists two update packages (2016 and 2019) install both, and confirm in Check product and patch installation status that each server and product shows the new build with no 'upgrade required' state; run PSConfig after the update if the status page says it is needed. | | the attacker has a low-privilege account or reaches an anonymous site | The anonymous route works only on a farm that lacks the 9 Jun 2026 fix. List who can sign in: Central Administration > Application Management > Manage web applications, then review user policy and site permissions. Any account with a valid sign-in on an exposed site meets the condition. For the anonymous route, in Central Administration > Application Management > Manage web applications > select the web application > Authentication Providers > select the zone, check 'Enable anonymous access', then open Site Settings > Site permissions > Anonymous Access on a site. Previdian says the 9 Jun 2026 fix closes the anonymous route but names no update; the fixed builds in the versions table are later than the June 2026 SharePoint security update builds, so a farm at or above them has that fix too. With anonymous access off, the delivery route needs a signed-in account. Network reach: check which IP addresses and ports the web application binds to in IIS Manager > Sites, and whether the firewall or reverse proxy publishes it to the internet. | ## Affected and fixed versions | Product line | Affected builds | First fixed build | Update packages | | --- | --- | --- | --- | | SharePoint Server 2016 (Enterprise Server 2016 uses the same updates) | Earlier than 16.0.5565.1001 | 16.0.5565.1001 | KB5002905 and KB5002906 | | SharePoint Server 2019 | Earlier than 16.0.10417.20198 | 16.0.10417.20198 | KB5002894 and KB5002896 | | SharePoint Server Subscription Edition | Earlier than 16.0.19725.20522 | 16.0.19725.20522 | KB5002893 | Microsoft says to install every update package offered for the installed product, not just one. The updates may need a restart. Microsoft's FAQ says the same KB applies to SharePoint Server 2016 and SharePoint Enterprise Server 2016. ## If the fix cannot go in yet - Install the August 2026 SharePoint update for each product on every farm server (KB5002905 and KB5002906, KB5002894 and KB5002896, or KB5002893). Cost: a restart may be needed, and the farm configuration step must finish on every server. This is the only fix, and Microsoft's records list no workaround. - If you cannot patch yet, turn off anonymous viewing on web applications that do not need it (Previdian). Cost: public visitors must then sign in. This closes only the anonymous route Previdian saw, not the authenticated flaw. - Limit SharePoint exposure to the internet, for example by placing it behind a VPN or an authenticating reverse proxy. Cost: remote users and partners lose direct access. An attacker with a valid low-privilege account on the internal network can still reach it. ## What to look for - Previdian hunting guidance: search web logs for POST requests to /_layouts/15/AddGallery.aspx and /_layouts/15/designgallery.aspx, including paths with a repeated /_layouts/ or /_layouts/15/ prefix. The captured requests used the query string job=all&DisplayMode=Edit, sent no Cookie or Authorization header, and came in pairs of bodies of 7,834 and 535,404 bytes. - Files to look for (Previdian): a web shell at /_layouts/15/sphealth.aspx, and assemblies wt3k3sij.dll (391,680 bytes, SHA-256 d3faa4b443d98f272363f3484a5e6a9bab90979086aa2d31a1694c1dc8178742) and 24e5mo4s.dll (4,096 bytes, SHA-256 a151a8fc193a96aac480fa749547b57c0f33116cf2cb8c82b6aa716c3c47f4b1). Previdian did not see wt3k3sij.dll written to disk on its honeypot. - Network indicator: source address 169.150.248[.]21 (AS212238, Datacamp Limited), seen on 24 Sep 2026 near 12:00 UTC. The URL http://asdf/_controltemplates/15/AclEditor.ascx appeared in a form field and is not command-and-control infrastructure. - Hunting only for the older ToolPane.aspx pattern from earlier SharePoint attacks misses this traffic. CISA's KEV entry marks forensic triage as required; its required action cites CISA's BOD 26-04. ## Timeline | Date | Event | Source | | --- | --- | --- | | 9 Jun 2026 | Per Viettel Cyber Security as cited by Previdian, the separate anonymous-delivery weakness is fixed. This is the June 2026 Patch Tuesday; Microsoft's June SharePoint security updates were KB5002873 (Subscription Edition, 16.0.19725.20384), KB5002874 and KB5002876 (2019, 16.0.10417.20153) and KB5002880 and KB5002881 (2016, 16.0.5556.1005). No source ties this weakness to one of those KBs. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Jun) | | 11 Aug 2026 | Microsoft publishes the advisory (revision 1.0) as a spoofing issue rated CVSS 3.1 6.5, with the security updates; NVD publishes the record. Microsoft credits khoadha. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Aug) | | 27 Aug 2026 | Microsoft revision 1.1 changes the impact, title and FAQs to remote code execution and the CVSS 3.1 base score to 8.8, marked informational. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Aug) | | 24 Sep 2026 | Previdian captures 12 POST requests from one source on a honeypot near 12:00 UTC and publishes its report. The requests are attempts; Previdian did not recover the final payload or show a compromise. CISA's SSVC entry on the NVD record carries the same date and marks exploitation active. | [source](https://blog.previdian.com/cve-2026-65660-previdian-observes-two-stage-sharepoint-exploitation-attempts/) | | 25 Sep 2026 | Microsoft revision 1.2 states it had reliable evidence of observed attacks, and CISA adds the CVE to KEV with a due date of 28 Sep 2026. KEV lists ransomware use as Unknown. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Aug) | Exploited 44 days after disclosure: disclosed 11 Aug 2026, first exploitation reported 24 Sep 2026. ## Exploitation reported so far Previdian's honeypot received 12 requests from one source address, carrying two payloads per path: one to disable a .NET deserialization safeguard, then a larger one carrying an assembly loader. Previdian classed the traffic as an attempted pre-authentication chain. ## Why runtime reachability settles it A build check says whether the farm is patched, but exposure also depends on who can reach the server and whether any site allows anonymous viewing, which are facts about the running farm. The first row of the checks table settles the patch state and the second row settles who can reach the flaw. KEV and Microsoft confirm attacks, so a reachable unpatched farm is urgent. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-68820: Microsoft Windows AFD for WinSock](/ledger/cve-2026-68820/) - [CVE-2026-42018: JFrog Artifactory](/ledger/cve-2026-42018/) - [CVE-2026-41709: Broadcom VMware ESX, VMware ESX insufficient logging](/ledger/cve-2026-41709/) ## Sources - CVE-2026-65660: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-65660 (read 19 Nov 2026) - CVE-2026-65660: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-65660 (read 19 Nov 2026) - CVE-2026-65660: Microsoft advisory (MSRC API record: description, FAQs, severity): https://api.msrc.microsoft.com/sug/v2.0/en-US/vulnerability/CVE-2026-65660 (read 19 Nov 2026) - CVE-2026-65660: Microsoft advisory (MSRC August 2026 CVRF: fixed builds, KBs, revision history): https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Aug (read 19 Nov 2026) - CVE-2026-65660: Previdian research on exploitation attempts: https://blog.previdian.com/cve-2026-65660-previdian-observes-two-stage-sharepoint-exploitation-attempts/ (read 19 Nov 2026) - CVE-2026-65660: Microsoft advisory (MSRC June 2026 CVRF: SharePoint KBs and builds of 9 Jun 2026): https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Jun (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-65660/ --- # CVE-2026-68820: Microsoft Windows AFD for WinSock, is it exploitable where you run it Ledger entry, disclosed 11 Aug 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-68820 affects Microsoft Windows AFD for WinSock and was disclosed 11 Aug 2026 with a CVSS score of 7.0. Exposure needs these conditions: the host build is below the fixed build in the versions table; an attacker already runs code on the host as a local user. CISA added it to the Known Exploited Vulnerabilities catalog on 11 Aug 2026. ## What the flaw is CVE-2026-68820 is a use-after-free (CWE-416) in the Windows Ancillary Function Driver for WinSock, the kernel driver afd.sys that backs Windows sockets. Microsoft says a locally authenticated attacker can run a specially crafted application on an affected system, win a race condition, and gain SYSTEM privileges. No user interaction is needed. Microsoft rates the attack complexity High because the attacker must win a race condition. The flaw is local only: it does not give initial access. It is an escalation step for an attacker who already runs code as a normal user. The CVSS 3.1 base score is 7.0. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the host build is below the fixed build in the versions table | Run winver, or in PowerShell [System.Environment]::OSVersion.Version, or cmd /c ver. Compare the full build and revision (for example 10.0.19045.7663) with the versions table. Get-HotFix -Id KB5120249 (use the KB for your release) or wmic qfe list brief also works; a later cumulative update carries the fix, so compare builds when the KB is absent. Hosts on the hotpatch baseline report the earlier build in the table's fixed column, and are fixed. | | an attacker already runs code on the host as a local user | This is a property of your environment (CVSS AV:L, PR:L: a local, authorized account). List who can sign in locally or by Remote Desktop: net localgroup "Remote Desktop Users", net localgroup Users and, on a shared host, query user to see live sessions. Hosts where many people run code (Remote Desktop session hosts, VDI pools, jump hosts, build agents) come first. Service accounts: Get-CimInstance Win32_Service \| Where-Object { $_.StartName -notmatch 'LocalSystem\|NT AUTHORITY' } \| Select-Object Name, StartName lists services that run as ordinary accounts, and code execution in any of them is a local foothold. Application control state: Get-CimInstance -ClassName Win32_DeviceGuard -Namespace root\Microsoft\Windows\DeviceGuard shows CodeIntegrityPolicyEnforcementStatus (1 is audit only, 2 is enforced), and Get-AppLockerPolicy -Effective lists AppLocker rules; a host with no enforced policy lets any local user run a crafted application. | ## Affected and fixed versions | Windows release | Affected builds (NVD threshold) | Fixed builds as Microsoft lists them | Update (11 Aug 2026) | | --- | --- | --- | --- | | Windows Server 2012 (and Server Core) | Earlier than 6.2.9200.26280 | 6.2.9200.26280 | KB5120386 | | Windows Server 2012 R2 (and Server Core) | Earlier than 6.3.9600.23338 | 6.3.9600.23338 | KB5120385 | | Windows 10 Version 1607 and Windows Server 2016 (and Server Core) | Earlier than 10.0.14393.9418 | 10.0.14393.9418 | KB5120418 | | Windows 10 Version 1809 and Windows Server 2019 (and Server Core) | Earlier than 10.0.17763.9121 | 10.0.17763.9121 | KB5120238 | | Windows 10 Version 21H2 (32-bit, ARM64, x64) | Earlier than 10.0.19044.7663 | 10.0.19044.7663 | KB5120249 | | Windows 10 Version 22H2 (32-bit, ARM64, x64) | Earlier than 10.0.19045.7663 | 10.0.19045.7663 | KB5120249 | | Windows Server 2022 (and Server Core) | Earlier than 10.0.20348.5499 | 10.0.20348.5440 (KB5120229) or 10.0.20348.5499 (KB5120242) | KB5120229, KB5120242 | | Windows 11 Version 23H2 (ARM64, x64) | Earlier than 10.0.22631.7517 | 10.0.22631.7517 | KB5120240 | | Windows 11 Version 24H2 (ARM64, x64) | Earlier than 10.0.26100.9168 | 10.0.26100.9106 (KB5120994) or 10.0.26100.9168 (KB5121003) | KB5120994, KB5121003 | | Windows 11 Version 25H2 (ARM64, x64) | Earlier than 10.0.26200.9168 | 10.0.26200.9106 (KB5120994) or 10.0.26200.9168 (KB5121003) | KB5120994, KB5121003 | | Windows 11 Version 26H1 (ARM64, x64) | Earlier than 10.0.28000.2704 | 10.0.28000.2704 | KB5121000 | | Windows Server 2025 (and Server Core) | Earlier than 10.0.26100.33296 | 10.0.26100.33222 (KB5120228) or 10.0.26100.33296 (KB5120233) | KB5120228, KB5120233 | Where Microsoft lists two KBs for one release (Server 2022, Windows 11 24H2 and 25H2, Server 2025), its release data marks the earlier build's KB as a Security Hotpatch Update and the later build's KB as the standard Security Update. Each carries the fix and each has its own fixed build in the table. NVD sets its affected threshold at the later build, so a hotpatched host on the earlier build still shows as below NVD's threshold even though Microsoft lists that build as fixed. The other releases list one KB. Microsoft's August 2026 release data lists 13 KB articles for this CVE in all. Windows 11 24H2 (build 26100.9168) and Windows Server 2025 (build 26100.33296) share the 26100 base build but ship on separate servicing streams, so their revision numbers differ; Microsoft's August 2026 release data lists each build against its own KB, and the table copies them. ## If the fix cannot go in yet - Install the August 2026 cumulative update, or a later one, for the release (see the versions table). Cost: a restart. This is the fix. - Microsoft lists no workaround. Because the attacker must already run code locally, patch the hosts where that is routine first: Remote Desktop session hosts, VDI pools, jump hosts and build agents. Cost: an earlier restart window for those hosts. ## What to look for - Microsoft and CISA publish no log signature, file name or network indicator for this flaw, and CISA marks forensic triage as not required. The one artifact tied to the bug is the driver. Microsoft says the attacker must win a race, so an attempt that loses can leave a kernel crash. In the System log, Microsoft-Windows-WER-SystemErrorReporting event 1001 records each BugCheck. Open the dump (C:\Windows\MEMORY.DMP or C:\Windows\Minidump) in WinDbg and run !analyze -v: a faulting module or stack that names afd.sys on a host below the fixed build is a lead. The same crash on a patched host is a different fault. Start with hosts where several users run code. ## Timeline | Date | Event | Source | | --- | --- | --- | | 11 Aug 2026 | Microsoft publishes the advisory (revision 1.0, "Information published") and the fix, marked Exploited: Yes and Publicly Disclosed: No, with David Driker and Moshe Marelus with Check Point credited. NVD publishes the record. | [source](https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Aug) | | 11 Aug 2026 | CISA adds the CVE to KEV on the same day, with a due date of 25 Aug 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-68820) | Exploited on the day of disclosure, 11 Aug 2026. ## Exploitation reported so far Microsoft's first release of the advisory marks the flaw as exploited, and CISA lists it in KEV. KEV lists ransomware campaign use as Unknown. Neither Microsoft nor CISA names the actors or victims, and no report on the exploitation is cited. ## Why attack-path analysis settles it Exposure needs an unpatched build and an attacker who already runs code locally on that host. The build check settles the first. Because the flaw is local, an attack-path view of how an attacker gets local code on unpatched hosts decides real priority, and KEV shows it is already used. ## Related pages - [CVE-2026-41709: Broadcom VMware ESX, VMware ESX insufficient logging](/ledger/cve-2026-41709/) - [CVE-2026-65660: Microsoft SharePoint Server, SharePoint code injection](/ledger/cve-2026-65660/) - [CVE-2026-42016: JFrog Artifactory](/ledger/cve-2026-42016/) ## Sources - CVE-2026-68820: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-68820 (read 19 Nov 2026) - CVE-2026-68820: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-68820 (read 19 Nov 2026) - CVE-2026-68820: Microsoft advisory (MSRC CVRF, August 2026 release: fixed builds, KBs, Exploited: Yes, FAQs, acknowledgments): https://api.msrc.microsoft.com/cvrf/v3.0/cvrf/2026-Aug (read 19 Nov 2026) - CVE-2026-68820: Microsoft advisory (MSRC API record): https://api.msrc.microsoft.com/sug/v2.0/en-US/vulnerability/CVE-2026-68820 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-68820/ --- # CVE-2026-41709: Broadcom VMware ESX, is it exploitable where you run it Ledger entry, disclosed 29 Jul 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-41709 affects Broadcom VMware ESX and was disclosed 29 Jul 2026 with a CVSS score of 2.7. Exposure needs these conditions: the host runs ESX 9.0, 8.0, VMware Cloud Foundation 5.x or Telco Cloud Platform 5.x below its fixed build (ESX 9.1 is not affected); an attacker holds administrator rights and can reach the management interface. ## What the flaw is VMware ESX does not write a log entry for some administrative operations (CWE-778, insufficient logging). Broadcom says a malicious administrator could use this to perform certain operations without them being logged. The advisory does not name the operations. The flaw gives an attacker no new access. It needs ESX administrator rights already, and it changes no data or service by itself. What the attacker gains is a quieter trail: actions taken with those rights are missing from the host's logs, so an incident responder who relies on ESX logs may miss them. Broadcom rates it 2.7 (CVSS base score), Low (integrity impact low, nothing else). It is one of five CVEs fixed in advisory VMSA-2026-0006, which Broadcom rates Critical overall. The advisory's Critical CVEs are two 9.8 vCenter flaws (CVE-2026-59309, an authentication bypass, and CVE-2026-59310, a directory traversal) and a 9.3 ESX VMXNET3 out-of-bounds write (CVE-2026-47876). CVE-2026-41703 is rated Important (7.6). Broadcom credits Ian Barton of CrowdStrike with the report. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the host runs ESX 9.0, 8.0, VMware Cloud Foundation 5.x or Telco Cloud Platform 5.x below its fixed build (ESX 9.1 is not affected) | On the host (SSH or ESXi Shell) run esxcli system version get and read the Build field, or run vmware -vl. In the Host Client the same build shows under Help > About. Compare with 25595025 (9.0) or 25429389 (8.0). A 9.1 host is not affected: Broadcom lists the 9.1.0.0 release, build 25370933, as the fixed version, so every 9.1 build has the fix. To see the installed image profile name, run esxcli software profile get. For VMware Cloud Foundation 5.x or Telco Cloud Platform, read the product version in SDDC Manager or the Telco Cloud Platform documentation, then confirm the build on each host with esxcli system version get: 25429389 or higher in the 8.0 U3 train (the versions table explains why), which is also the Telco Cloud Platform fix, ESXi80U3j-25429389. | | an attacker holds administrator rights and can reach the management interface | List who holds them: esxcli system permission list on the host, and in vCenter the host's Permissions tab. Review local accounts and any Active Directory group mapped to the Administrator role, and the vSphere SSO groups that reach the host. Check lockdown mode with vim-cmd -U dcui vimsvc/auth/lockdown_is_enabled, and review the host firewall with esxcli network firewall ruleset list for services such as sshServer and vSphere Web Access, and the allowed IP list on each. | ## Affected and fixed versions | Product line | Component and version | Fixed version | Workaround | | --- | --- | --- | --- | | VMware Cloud Foundation, VMware vSphere Foundation | ESX 9.1.x.x | ESXi-9.1.0.0-25370933, the 9.1.0.0 release itself; no 9.1 build is affected | None | | VMware Cloud Foundation, VMware vSphere Foundation | ESX 9.0.x.x | ESXi-9.0.2.0100-25595025 | None | | VMware ESX | 8.0 | ESXi80U3j-25429389 | None | | VMware Cloud Foundation | ESX 5.x | VMware Cloud Foundation 5.2.4, a VCF release; the matrix names the release, not an ESX build | None | | VMware Telco Cloud Platform | ESX 5.0.x and 5.1.x | ESXi80U3j-25429389 (8.0 U3j), per KB449886 | None | Broadcom's response matrix (VMSA-2026-0006, revision 2, 19 Aug 2026) gives one fixed build per line and treats earlier builds as affected. Build numbers belong to separate trains, so the 9.1 fix (25370933) is lower than the 9.0 fix (25595025). Telco Cloud Platform 5.0.x and 5.1.x point to KB449886, which lists ESXi80U3j-25429389. On a VCF 5.x host, Broadcom's async patch article (KB344935, reached from the matrix link KB88287) lists ESX 8.0 U3j as build 8.0.3-25429389, the build the matrix gives for standalone ESX 8.0, so a host at or above 25429389 in the 8.0 U3 train (U3k is 25595708) carries the standalone fix. The matrix lists ESX 7.0 for the Critical sibling CVE-2026-47876 (contact Broadcom Support if you have an extended support contract) but has no 7.0 row for CVE-2026-41709, so Broadcom states no fix and no status for 7.0 here; owners of 7.0 hosts should ask Broadcom Support and, until they hear, treat those hosts as unpatched and limit administrator access. Patches are cumulative: ESXi-9.1.0.0200-25557999 and ESXi80U3k-25595708, which fix CVE-2026-47876, also cover this CVE. On 9.0, ESXi-9.0.2.0100-25595025 is the fix for both CVEs. ## If the fix cannot go in yet - Broadcom lists no workaround. Apply the fixed build for your line from the versions table. Cost: the host must enter maintenance mode and reboot, so plan vMotion or a rolling cluster update. - Until the patch is in, cut the number of people and systems that hold ESX administrator rights and restrict who can reach the host management interface. This is general hardening, not a Broadcom instruction. Cost: administrators lose direct host access and must go through vCenter or a jump host. It does not remove the flaw: an administrator you still allow can use it. - Forward ESX syslog to a remote collector (esxcli system syslog config set --loghost=, then esxcli system syslog reload). This is general hardening, not a Broadcom instruction. It protects the entries that are written from later deletion on the host, but it cannot capture the operations this flaw leaves unlogged. It does not decide exposure, so it is not a condition above. ## What to look for - Broadcom does not name which operations go unlogged, and no vendor or CISA source publishes a log string, event ID or file path for this CVE, so the host logs hold nothing to search for. - Compensating check for unlogged administrator actions: compare the host's own record (/var/log/hostd.log, /var/log/auth.log and /var/log/shell.log) with the central record. In the vSphere Client open the host, then Monitor > Tasks and Events and filter by user and time, or run Get-VIEvent -Entity in PowerCLI. A change visible in vCenter's task history or in the host's configuration but absent from the host logs for the same account and time is the pattern to investigate. This is general practice, not a Broadcom instruction. - Review the accounts that hold the Administrator role on each host (esxcli system permission list) and the SSH and ESXi Shell logins in /var/log/auth.log for the weeks before patching, since the flaw needs administrator rights to be used. - CISA's KEV entry sets a due date of 11 Nov 2026. Its required action reads: apply mitigations in accordance with vendor instructions, and discontinue use of the product if mitigations are unavailable; the KEV text cites CISA's BOD 26-04 for the patching timeline. ## Timeline | Date | Event | Source | | --- | --- | --- | | 29 Jul 2026 | Broadcom publishes advisory VMSA-2026-0006 with fixed builds for CVE-2026-41709 and four other CVEs. | [source](https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017) | | 30 Jul 2026 | NVD publishes the record, with Broadcom's CVSS 2.7 (Low) as its rating. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-41709) | | 19 Aug 2026 | Broadcom updates the advisory (revision 2); the update date is listed as 19 Aug 2026. | [source](https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017) | | 28 Oct 2026 | CISA adds the CVE to KEV with a due date of 11 Nov 2026. CISA's SSVC entry on the NVD record, stamped the same day, lists exploitation as active, automatable as no and technical impact as partial. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-41709) | The earliest dated mark of exploitation is CISA's, on 28 Oct 2026, 91 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far The KEV listing came 91 days after Broadcom's advisory. KEV lists ransomware campaign use as Unknown. No public report names the actor, the operations that were hidden or the attack that used it. Broadcom's advisory does not mention exploitation. A low CVSS score does not make this harmless in an incident. An unlogged administrator action is useful to someone who has already taken over an ESX administrator account, because it hides what they did from the host's logs. Neither CISA nor Broadcom describes the attack that used it. ## Why known-exploited lists settle it The CVSS score of 2.7 and the high privilege requirement would put this at the back of a queue ordered by CVSS score, and the flaw leaves no log trail to search. The KEV listing and CISA's SSVC entry (exploitation: active) are the only evidence of use, so a known-exploited check is what raises it. The build number from esxcli system version get then tells you which hosts to patch. See [known-exploited lists](/methods/known-exploited/) for what the method proves and misses. ## Related pages - [CVE-2026-42016: JFrog Artifactory](/ledger/cve-2026-42016/) - [CVE-2026-68820: Microsoft Windows AFD for WinSock](/ledger/cve-2026-68820/) - [CVE-2026-16723: Alibaba fastjson 1.x](/ledger/cve-2026-16723/) ## Sources - CVE-2026-41709: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-41709 (read 19 Nov 2026) - CVE-2026-41709: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-41709 (read 19 Nov 2026) - CVE-2026-41709: Broadcom security advisory VMSA-2026-0006 (vendor advisory, response matrix): https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017 (read 19 Nov 2026) - CVE-2026-41709: Broadcom KB449886, VMware Telco Cloud response to VMSA-2026-0006: https://knowledge.broadcom.com/external/article/449886 (read 19 Nov 2026) - CVE-2026-41709: Broadcom KB344935, Async Patch listing of ESX builds for VCF 5.2.x: https://knowledge.broadcom.com/external/article?legacyId=88287 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-41709/ --- # CVE-2026-42016: JFrog Artifactory, is it exploitable where you run it Ledger entry, disclosed 27 Jul 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-42016 affects JFrog Artifactory and was disclosed 27 Jul 2026 with a CVSS score of 8.8. Exposure needs the instance is self-hosted Artifactory on a build below 7.133.11, or on the 7.111, 7.117 or 7.125 train below the target build in the versions table (JFrog Cloud was fixed by JFrog). In the CISA KEV catalog since 11 Sep 2026. ## What the flaw is Artifactory's Access service issues and accepts tokens. When a caller presents a token, Artifactory confirms that the signature and the issuer are valid, but it does not enforce the scope the token was meant to carry. JFrog classes this as incorrect authorization (CWE-863). An attacker who holds any valid low-privileged token can ask the token endpoint for a new token with administrator authority. The new token keeps the original username but carries admin rights. The attacker then controls users, repositories, configuration and stored credentials. NVD rates it 8.8 (CVSS base score; network, low complexity, low privileges, no user interaction). On its own the bug needs a low-privileged token. CVE-2026-42018 can supply one without any login, which is how attackers used the pair in the wild. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the instance is self-hosted Artifactory on a build below 7.133.11, or on the 7.111, 7.117 or 7.125 train below the target build in the versions table (JFrog Cloud was fixed by JFrog) | Run GET /artifactory/api/system/version and read the version field; JFrog documents this call. Below 7.133.11 is affected. On the 7.111, 7.117 and 7.125 trains JFrog lists no fixed build for this CVE, so use the target build for the train in the versions table. Self-hosted means you run the Artifactory service yourself (the artifactory service under systemctl, or the container or Helm release). | | a low-privileged token can be obtained and the Access token endpoint is reachable from the attacker's network position | Any valid user token qualifies. Check whether anonymous access is enabled (Administration, Security, General), and whether the instance is also open to CVE-2026-42018, which returns an anonymous token even with anonymous access disabled (it has its own ledger entry). From a host where an attacker could sit, confirm the paths under /access/api/v1/ answer through your reverse proxy or load balancer. Wiz saw the exploit use POST /access/api/v1/tokens. | ## Affected and fixed versions | Release train | Affected (JFrog and NVD) | Minimum JFrog lists | Target build for the chain | | --- | --- | --- | --- | | 7.111.x | Below 7.133.11, so all builds | None listed | 7.111.21 | | 7.117.x | Below 7.133.11, so all builds | None listed | 7.117.28 | | 7.125.x | Below 7.133.11, so all builds | None listed | 7.125.20 | | 7.133.x | 7.133.0 to 7.133.10 | 7.133.11 | 7.133.29 | | 7.146.x | Not affected (above the bound) | Not applicable | 7.146.38 | | 7.161.x | Not affected (above the bound) | Not applicable | 7.161.20 | JFrog's table for this CVE says only: Artifactory, affected versions below 7.133.11, patched version 7.133.11; NVD uses the same bound. It publishes no 42016-specific build for the older trains. The target column is the set Wiz tells teams to move to for the CVE-2026-42016 and CVE-2026-42018 chain, and it matches JFrog's fixed builds for CVE-2026-82329 (28 Aug 2026), so one upgrade to that build closes all three. The CVE-2026-42018 entry uses the same target column. ## If the fix cannot go in yet - Upgrade to the target build for your train in the versions table (7.133.11 is the lowest build JFrog lists). JFrog gives no workaround for this CVE, so the upgrade is the fix. Cost: a restart of the service and the usual upgrade testing. - Limit the Access paths and the Artifactory UI to trusted networks (Wiz). Cost: remote developers, CI systems and external consumers lose access unless they come through the allowed path. - Treat the instance as compromised if it was exposed while vulnerable (Wiz): an upgrade does not remove admin users, Groovy plugins or stolen Access signing keys. Review admin users and the plugins directory, and reset the Access token certificate if you find signs of entry. Cost: a reset invalidates tokens and needs a planned window. ## What to look for - Look for a low-privilege identity or the internal anonymous identity minting tokens at POST /access/api/v1/tokens or POST /artifactory/api/security/token (Wiz). - Sample query for admin actions by the escalated anonymous token. After a successful exchange the token keeps the anonymous username but carries admin authority, so later requests show an actor of token:anonymous. In the request log (artifactory-request.log under $JFROG_HOME/artifactory/var/log/, pipe-separated fields) run: grep -F 'token:anonymous' artifactory-request.log* | grep -E '\|(PUT|POST|DELETE|PATCH)\|' | grep -E '/api/security/(users|token)|/access/api/(v1/tokens|ui/users)|/api/plugins'. Any hit with a 200 or 201 status is a candidate; adjust the field match if your log layout differs. - Look for a low-privilege identity listing users at GET /access/api/v1/users or GET /artifactory/api/security/users/, and any low-privilege identity reading or writing /artifactory/api/plugins (Wiz). - Persistence to look for: new admin accounts created by PUT /api/security/users/ or /access/api/ui/users/ (Wiz saw names such as jfrog-distribution, jfrog-insight, repo-service, backup-service, migration-tool, ldap_admin and 0xterror), and Groovy files under $JFROG_HOME/var/etc/artifactory/plugins/. Wiz recovered plugins named rce.groovy, cmd.groovy, jwtToken.groovy, uploadFile.groovy, JFrogImage.groovy, metrics.groovy and httpSession.groovy. - Post-exploitation indicators Wiz saw with the chain: payload downloads from hxxp://log.gitclone[.]org:45678/smtp and hxxp://3.88.162[.]79:36789/smtp, C2 64.207.232[.]6:8443, payload SHA1 513a907b69edffc3cb77a494da395178d21ef9bd (file /tmp/.z), and actor IPs 149.102.229[.]150 and 121.123.141[.]155. ## Timeline | Date | Event | Source | | --- | --- | --- | | 27 Jul 2026 | JFrog publishes the advisory and NVD publishes the record for CVE-2026-42016. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 12 Aug 2026 | JFrog publishes CVE-2026-42018, the bug that supplies the low-privileged token. | [source](https://docs.jfrog.com/releases/docs/jfrog-security-advisories) | | 15 Aug 2026 | Start of the window in which Wiz observed the CVE-2026-42018 and CVE-2026-42016 chain in the wild (15 Aug to 8 Sep 2026). | [source](https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201) | | 10 Sep 2026 | Wiz publishes its report of in-the-wild exploitation. | [source](https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201) | | 11 Sep 2026 | CISA adds the CVE to KEV. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42016) | | 17 Sep 2026 | Wiz adds the malicious Groovy plugin section and more indicators to its report. | [source](https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201) | | 25 Sep 2026 | CISA KEV due date for federal agencies. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42016) | Exploited 19 days after disclosure: disclosed 27 Jul 2026, first exploitation reported 15 Aug 2026. ## Exploitation reported so far Wiz Research says attackers chained CVE-2026-42018 and CVE-2026-42016 against self-hosted Artifactory between 15 Aug and 8 Sep 2026, and knew of no earlier public report. Several actors were involved, and some went from the first request to a new admin account in under five minutes. Follow-on activity included persistent admin accounts, Groovy plugins, shell commands, and a custom Rust backdoor dropped under /tmp or /dev/shm style paths. Wiz measured that 59% of organizations running Artifactory still had a vulnerable instance for this CVE six weeks after it was first published on 27 Jul 2026, down from 67% at publication. Its figure for CVE-2026-42018 (62% four weeks after its 12 Aug 2026 publication) is a different CVE measured from a different date. ## Why attack-path analysis settles it The bug has two inputs: an affected build and a low-privileged token. The version call settles the first, but only the path from an untrusted network to a token source (the anonymous token of CVE-2026-42018, open anonymous access or a stolen user token) settles whether an attacker can reach the second, so attack-path analysis fits. Because the pair is known exploited, treat a vulnerable build with a reachable Access path as already compromised. ## Related pages - [CVE-2026-16723: Alibaba fastjson 1.x](/ledger/cve-2026-16723/) - [CVE-2026-41709: Broadcom VMware ESX, VMware ESX insufficient logging](/ledger/cve-2026-41709/) - [CVE-2026-59822: BerriAI LiteLLM, LiteLLM MCP endpoint auth bypass](/ledger/cve-2026-59822/) ## Sources - CVE-2026-42016: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-42016 (read 19 Nov 2026) - CVE-2026-42016: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-42016 (read 19 Nov 2026) - CVE-2026-42016: JFrog security advisories (vendor advisory, CVE-2026-42016 section): https://docs.jfrog.com/releases/docs/jfrog-security-advisories (read 19 Nov 2026) - CVE-2026-42016: Wiz Research report, 10 Sep 2026, updated 17 Sep 2026 (the URL as Wiz publishes it, which NVD lists too): https://www.wiz.io/blog/artifactory-under-attack-in-the-wild-exploitation-of-cve-2026-42016-cve-2026-4201 (read 19 Nov 2026) - CVE-2026-42016: JFrog documentation: Get Artifactory Version API: https://docs.jfrog.com/administration/reference/getartifactoryversion (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-42016/ --- # CVE-2026-16723: Alibaba fastjson 1.x, is it exploitable where you run it Ledger entry, disclosed 21 Jul 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-16723 affects Alibaba fastjson 1.x and was disclosed 21 Jul 2026 with a CVSS score of 9.0. Exposure needs these conditions: the application ships fastjson 1.2.68 to 1.2.83 and is not a noneautotype build; it runs as a Spring Boot executable fat-jar with SafeMode off and passes untrusted JSON to JSON.parse or JSON.parseObject. ## What the flaw is fastjson is a widely used Java JSON library. CVE-2026-16723 is a remote code execution flaw (CWE-502, deserialization of untrusted data) in the 1.x line. It can be triggered under the stock default configuration: AutoType is off and SafeMode is off, and no AutoType enablement or classpath gadget is needed. Per the 1.2.84 notes, a type name that contains URL-special characters could reach resource probing and the class loader, and the fix rejects such names before that point. The advisory names one deployment prerequisite: the application runs as a Spring Boot executable fat-jar (started with java -jar). It says the flaw is reachable through JSON.parse, JSON.parseObject(String) and JSON.parseObject(String, Class). Naming a target class is not a mitigation, because an attacker can nest the payload inside an Object or Map field of that class. An attacker who can send JSON to such a parse call can run code in the application. The advisory is dated 21 Jul 2026, the disclosure date; NVD published the record on 23 Jul 2026. The CNA rates it 9.0 (CVSS 3.1 base score): network, high attack complexity, no privileges, scope changed. NVD lists the record as Deferred and has added no CVSS score of its own. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the application ships fastjson 1.2.68 to 1.2.83 and is not a noneautotype build | Maven: mvn dependency:tree -Dincludes=com.alibaba:fastjson prints a line such as `+- com.alibaba:fastjson:jar:1.2.83:compile` for an affected build and `+- com.alibaba:fastjson:jar:1.2.83_noneautotype:compile` for a noneautotype build. Gradle: gradle dependencies --configuration runtimeClasspath \| grep -i fastjson. On a built artifact: jar tf app.jar \| grep -i fastjson (a fat-jar lists BOOT-INF/lib/fastjson-1.2.x.jar). Also check your SBOM. A noneautotype build has the vulnerable code removed; a plain 1.2.83 does not. The advisory states that 1.2.60 and earlier lack the code path and that 1.2.84 and all fastjson2 versions are not affected. It gives no status for 1.2.61 to 1.2.67, so treat those as unconfirmed and move to 1.2.84. | | it runs as a Spring Boot executable fat-jar with SafeMode off and passes untrusted JSON to JSON.parse or JSON.parseObject | Fat-jar: `jar tf app.jar \| grep -m1 BOOT-INF` shows the Spring Boot nested layout (application classes in BOOT-INF/classes, dependencies in BOOT-INF/lib, per the Spring Boot executable jar documentation); on a running host `ps -o args= -p PID` shows java -jar app.jar. SafeMode off and AutoType off (both are the stock defaults): `ps -ef \| grep fastjson.parser.safeMode`, `unzip -p app.jar BOOT-INF/classes/fastjson.properties` for fastjson.parser.safeMode=true, and `grep -rn 'setSafeMode' src/`. Untrusted parse: `grep -rnE 'JSON\.parse(Object)?\(' src/` and trace each hit back to a request body, queue message, cache value or file an outsider can influence; naming a target class does not remove the exposure. A WAR in an external Tomcat or Jetty is outside the advisory's trigger condition. | ## Affected and fixed versions | Product line | Affected | First fixed or not affected | Notes | | --- | --- | --- | --- | | fastjson 1.x (com.alibaba:fastjson) | 1.2.68 to 1.2.83 | 1.2.84 (released 29 Jul 2026) | Versions 1.2.60 and earlier lack the vulnerable code path | | fastjson 1.x 1.2.61 to 1.2.67 | Not stated in the advisory | 1.2.84 | The advisory lists 1.2.68 as the first affected version and 1.2.60 as the last version without the path; it is silent between them, so upgrade to 1.2.84 | | fastjson 1.x noneautotype builds | Not affected | Not applicable | Vulnerable code removed at compile time; example artifact version 1.2.83_noneautotype | | fastjson2 | Not affected (all versions) | Not applicable | Type resolution was designed without the resource probing step | The advisory was dated 21 Jul 2026 and updated on 29 Jul 2026. The fastjson 1.x repository was archived by its owner on 29 Jul 2026 and is read-only. ## If the fix cannot go in yet - Upgrade to com.alibaba:fastjson:1.2.84. This is the fix. Cost: the 1.2.84 release also tightens AutoType handling, so an application that relies on an accept list prefix to cover base types such as ClassLoader, DataSource or RowSet will stop accepting them; only a full type name entry is an opt-in. - Turn on SafeMode with -Dfastjson.parser.safeMode=true, or ParserConfig.getGlobalInstance().setSafeMode(true), or fastjson.parser.safeMode=true in fastjson.properties (advisory, P0 action). Cost: SafeMode disables AutoType completely, so any code that depends on @type polymorphic deserialization stops working. - Switch to a noneautotype build (Maven example 1.2.83_noneautotype). Cost: the AutoType code is gone, so @type deserialization no longer works. - Migrate to fastjson2, which Alibaba rates as removing the issue under its default configuration (advisory P1 action). Cost: a code migration; the project publishes an upgrade guide and a compatibility package. - Avoid fat-jar deployment: the advisory's not-affected table lists plain jar, uber-jar and WAR deployments as not meeting the trigger condition. Cost: a change to the packaging and the run command, with its own testing. Prefer the upgrade. ## What to look for - Alibaba, NVD and CISA publish no indicators of compromise, log signatures or sample traffic for this CVE. The advisory's guidance is to find affected services and upgrade or enable SafeMode. - After finding the services, list the call sites that pass request data to JSON.parse or JSON.parseObject. - The 1.2.84 release notes say type names containing the characters : / or ! are now rejected before class loading. On a patched build, request bodies or WAF logs that carry an @type value with those characters therefore mark attempts to reach the flaw; Alibaba publishes no exception text for them. ## Timeline | Date | Event | Source | | --- | --- | --- | | 21 Jul 2026 | Alibaba publishes the security advisory on the fastjson2 wiki, crediting Kirill Firsov of FearsOff Cybersecurity. | [source](https://github.com/alibaba/fastjson2/wiki/Security-Advisory:-Remote-Code-Execution-in-fastjson-1.2.68%E2%80%931.2.83) | | 23 Jul 2026 | NVD publishes CVE-2026-16723 (source: Alibaba CNA). The CISA coordinator SSVC entry the same day lists exploitation none, automatable no, technical impact total. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-16723) | | 29 Jul 2026 | fastjson 1.2.84 is released with the fix and the advisory is updated. The fastjson 1.x repository is archived the same day. | [source](https://github.com/alibaba/fastjson/releases/tag/1.2.84) | No public report dates the start of exploitation. ## Exploitation reported so far CVE-2026-16723 is not in the CISA KEV catalog as of 18 Nov 2026, and the CISA coordinator SSVC entry on the NVD record, stamped 23 Jul 2026, rates exploitation as none. No use in the wild is published. Alibaba's advisory says the finder verified the flaw end to end on Spring Boot 2.x, 3.x and 4.x with JDK 8, 11, 17 and 21. The advisory and NVD cite no public proof of concept. ## Why static reachability settles it Every condition here is a fact in the build or the code: the fastjson version, fat-jar packaging, the SafeMode setting and a call path from untrusted input to the parser. Static reachability checks those without waiting for traffic. No one reports attacks yet, so a CVSS score alone would over- or under-state the risk for any one service. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-59822: BerriAI LiteLLM, LiteLLM MCP endpoint auth bypass](/ledger/cve-2026-59822/) - [CVE-2026-42016: JFrog Artifactory](/ledger/cve-2026-42016/) - [CVE-2026-9198: IBM Langflow OSS](/ledger/cve-2026-9198/) ## Sources - CVE-2026-16723: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-16723 (read 19 Nov 2026) - CVE-2026-16723: vendor advisory (fastjson2 wiki, dated 21 Jul 2026, updated 29 Jul 2026): https://github.com/alibaba/fastjson2/wiki/Security-Advisory:-Remote-Code-Execution-in-fastjson-1.2.68%E2%80%931.2.83 (read 19 Nov 2026) - CVE-2026-16723: vendor release notes for fastjson 1.2.84: https://github.com/alibaba/fastjson/releases/tag/1.2.84 (read 19 Nov 2026) - CVE-2026-16723: vendor SafeMode documentation: https://github.com/alibaba/fastjson/wiki/fastjson_safemode (read 19 Nov 2026) - CVE-2026-16723: Spring Boot executable jar documentation: https://docs.spring.io/spring-boot/specification/executable-jar/nested-jars.html (read 19 Nov 2026) - CVE-2026-16723: CISA KEV feed (version 2026.11.18, no entry for this CVE): https://www.cisa.gov/known-exploited-vulnerabilities-catalog (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-16723/ --- # CVE-2026-59822: BerriAI LiteLLM, is it exploitable where you run it Ledger entry, disclosed 8 Jul 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-59822 affects BerriAI LiteLLM and was disclosed 8 Jul 2026 with a CVSS score of 8.8. Exposure needs these conditions: the proxy runs LiteLLM earlier than 1.84.0; an MCP server in config.yaml sets allow_all_keys: true, and /mcp/ is reachable by the attacker. CISA added it to the Known Exploited Vulnerabilities catalog on 2 Sep 2026. Exploited 50 days after disclosure. ## What the flaw is LiteLLM is a proxy that routes calls to LLM providers. It can also front MCP (Model Context Protocol) tool servers. CVE-2026-59822 is an improper authentication flaw (CWE-287) in the proxy's MCP Streamable HTTP endpoint. The MCP auth handler supports an OAuth2 passthrough mode for upstream MCP servers. In affected versions, when LiteLLM key validation failed, the fallback path accepted the request anyway and replaced the failed result with an empty auth object. An attacker who can reach the proxy sends a made-up Authorization header and gets an authenticated MCP session without a valid LiteLLM key. From there the attacker can list and call the MCP tools the proxy has configured and reach the services behind them. The fix pull request says the exposure covers MCP servers marked allow_all_keys, because those servers are available to every caller. Wiz notes that a proxy often holds model-provider keys and cloud permissions, so the impact can extend well past the proxy. Ratings: the CNA (GitHub) rates it 8.8 under CVSS 4.0 (network, low complexity, no privileges, high confidentiality impact, low integrity impact). NVD's own CVSS 3.1 base score is 8.2 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:L/A:N); the two scales differ, so the numbers are not in conflict. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the proxy runs LiteLLM earlier than 1.84.0 | On the host run pip show litellm, or pip list \| grep -i litellm. For a container run docker exec CONTAINER pip show litellm. 1.84.0 is the first fixed release (GitHub release v1.84.0). GitHub's advisory range is every version below 1.84.0 and names no start; the versions table gives the first build that carries the OAuth2 fallback the advisory describes. | | an MCP server in config.yaml sets allow_all_keys: true, and /mcp/ is reachable by the attacker | Exposed servers: the fix pull request #26463 says the failed-key fallback returned an empty auth object whose tool access came from every server marked allow_all_keys: true, so any server with that setting is exposed on 1.83 and earlier. Run grep -n -A8 mcp_servers config.yaml (the --config path), then grep -n -B2 -A8 allow_all_keys config.yaml. A proxy with no MCP servers has no MCP tools to reach, and the pull request describes no access to a server without allow_all_keys. Reachability: from a network position an attacker could use, run curl -s -o /dev/null -w '%{http_code}\n' https://PROXY_HOST/mcp/. A 401 or 403 means the route is reachable and rejects anonymous callers; a timeout or block means the gateway already filters it. Also review gateway rules for /mcp/ and related MCP paths. Network control: check the bind address and firewall or security group rules for the proxy port with ss -ltnp \| grep litellm, then your firewall or cloud security group listing. The proxy should not accept connections from untrusted networks. | ## Affected and fixed versions | Product line | Affected | First fixed | Notes | | --- | --- | --- | --- | | LiteLLM (pip package litellm, proxy server) | 1.81.9 to 1.83.x | 1.84.0 (released 14 May 2026) | 1.81.9 is the first release with the OAuth2 fallback described in the advisory (nightly build of 7 Feb 2026, stable of 15 Feb 2026). The fix pull request #26463 was merged on 30 Apr 2026. | | LiteLLM (pip package litellm, proxy server) | Before 1.81.9 | 1.84.0 | Inside GitHub's range (below 1.84.0), but the OAuth2 fallback code is absent. The fix also closes a second route to the same empty auth object: the handler tests the full request URL, query string included, for the text .well-known. That test is present from 1.77.7 on, and the allow_all_keys server setting exists by 1.80.13. Upgrade rather than rely on the older code. | The GitHub release page for v1.84.0 states that the release contains breaking changes and points to the v1.84.0 release notes on docs.litellm.ai. Read them before an upgrade from an older line. ## If the fix cannot go in yet - Upgrade to LiteLLM 1.84.0 or later. This is the fix. Cost: the vendor flags 1.84.0 as a release with breaking changes, so test configuration and clients first. - If you cannot upgrade yet, disable MCP routes, or block /mcp/ and related MCP endpoints at the reverse proxy or API gateway (GitHub advisory workaround). Cost: every MCP client and agent that reaches tools through this proxy stops working until the route is restored. - Keep the proxy off untrusted networks, scope its cloud IAM permissions narrowly and restrict its outbound network access (Wiz guidance). Cost: tools or providers that rely on broad egress or permissions need explicit allow rules. - CISA's KEV entry gives this required action: apply mitigations in accordance with vendor instructions, follow the BOD 26-04 guidance the entry cites, and discontinue use of the product if mitigations are unavailable. The KEV due date was 16 Sep 2026. ## What to look for - Neither the vendor nor Wiz published a log signature for this CVE. The practical review: in proxy and gateway access logs, look for requests to /mcp/ and the MCP REST paths (for example /mcp-rest/) from addresses you do not know, especially ones carrying an Authorization header that does not match any key you issued, and tool listing or tool call activity that no known client explains. - Wiz says its honeypots saw single-character bearer tokens (Authorization: Bearer x), with GET /v1/models as its printed example. A burst of requests with tiny bearer values from one source, to /mcp/ or to a model route, matches that pattern. - Wiz guidance: watch the runtime for an AI server process spawning a shell or other unexpected child processes. - A separate LiteLLM flaw, CVE-2026-42271, has its own Wiz indicators (download server 185.62.1[.]8, staging directory /tmp/.dbus-cache/). Finding them means that other flaw was used and does not confirm this one. ## Timeline | Date | Event | Source | | --- | --- | --- | | 6 Feb 2026 | Commit 6a213fc (pull request #20602, OAuth2 passthrough for upstream MCP servers) adds the fallback that returns an empty auth object when LiteLLM key validation fails. It ships in the 1.81.9 nightly build of 7 Feb 2026. | [source](https://github.com/BerriAI/litellm/commit/6a213fc3bc) | | 30 Apr 2026 | The fix, pull request #26463 (tighten public-route detection and OAuth2 fallback gating), is merged. | [source](https://github.com/BerriAI/litellm/pull/26463) | | 14 May 2026 | LiteLLM v1.84.0 is released; it carries the fix. | [source](https://github.com/BerriAI/litellm/releases/tag/v1.84.0) | | 8 Jul 2026 | NVD publishes CVE-2026-59822 (source: GitHub). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-59822) | | 22 Jul 2026 | GitHub publishes the reviewed advisory GHSA-7488-6r32-c95q with CVSS 4.0 score 8.8. | [source](https://api.github.com/advisories?cve_id=CVE-2026-59822) | | 27 Aug 2026 | Wiz publishes its 90-day honeypot report and says it saw this bug exploited; the report gives no first day, so exploitation began on or before this date. | [source](https://www.wiz.io/blog/ai-infrastructure-honeypot) | | 2 Sep 2026 | CISA adds the CVE to KEV with a due date of 16 Sep 2026. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-59822) | Exploited 50 days after disclosure: disclosed 8 Jul 2026, first exploitation reported 27 Aug 2026. ## Exploitation reported so far Wiz Research published a report on 27 Aug 2026 covering 90 days of honeypot telemetry. It says it observed exploitation of CVE-2026-59822 against its LiteLLM honeypots, with requests that carried a one-character bearer token (Authorization: Bearer x). The one request Wiz prints is GET /v1/models, a model-listing route, not a request to /mcp/. Wiz describes the flaw as granting full MCP access to any bearer token and prints no /mcp/ request, so the report ties its honeypot traffic to this CVE by Wiz's own attribution. The report gives no first day for this bug, so the 50 days from the 8 Jul 2026 disclosure to the Wiz report shown on this page is a ceiling: use began on or before 27 Aug 2026 and may have begun earlier. The fix shipped in 1.84.0 on 14 May 2026, 55 days before the CVE record of 8 Jul 2026, and GitHub published the reviewed advisory on 22 Jul 2026. CISA added the CVE to KEV on 2 Sep 2026, six days after the Wiz report. KEV lists ransomware campaign use as Unknown. The same Wiz report ties a different LiteLLM flaw (CVE-2026-42271, a command injection in the MCP server test endpoints) to a Qilin-linked chain. That claim does not apply to CVE-2026-59822. ## Why runtime reachability settles it Exposure turns on three facts that only the running proxy shows together: a version below 1.84.0, an MCP server with allow_all_keys: true, and an MCP route an outsider can reach. KEV and Wiz confirm real attacks, so treat a reachable proxy as urgent. The checks table settles it row by row. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-9198: IBM Langflow OSS](/ledger/cve-2026-9198/) - [CVE-2026-16723: Alibaba fastjson 1.x](/ledger/cve-2026-16723/) - [CVE-2026-53266: Linux kernel](/ledger/cve-2026-53266/) ## Sources - CVE-2026-59822: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-59822 (read 19 Nov 2026) - CVE-2026-59822: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-59822 (read 19 Nov 2026) - CVE-2026-59822: vendor advisory (GitHub security advisory GHSA-7488-6r32-c95q, read through the GitHub advisory API record): https://github.com/BerriAI/litellm/security/advisories/GHSA-7488-6r32-c95q (read 19 Nov 2026) - CVE-2026-59822: GitHub advisory API record (publish date, CVSS 4.0, credits): https://api.github.com/advisories?cve_id=CVE-2026-59822 (read 19 Nov 2026) - CVE-2026-59822: Wiz Research honeypot report: https://www.wiz.io/blog/ai-infrastructure-honeypot (read 19 Nov 2026) - CVE-2026-59822: vendor release page: https://github.com/BerriAI/litellm/releases/tag/v1.84.0 (read 19 Nov 2026) - CVE-2026-59822: vendor fix pull request: https://github.com/BerriAI/litellm/pull/26463 (read 19 Nov 2026) - CVE-2026-59822: vendor MCP documentation: https://docs.litellm.ai/docs/mcp (read 19 Nov 2026) - CVE-2026-59822: vendor commit that adds the OAuth2 fallback (first shipped in 1.81.9): https://github.com/BerriAI/litellm/commit/6a213fc3bc (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-59822/ --- # CVE-2026-9198: IBM Langflow OSS, is it exploitable where you run it Ledger entry, disclosed 2 Jul 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-9198 affects IBM Langflow OSS and was disclosed 2 Jul 2026 with a CVSS score of 9.8. Exposure needs these conditions: Langflow OSS runs a version from 1.0.0 through 1.10.0; auto-login is enabled (LANGFLOW_AUTO_LOGIN set to true), which IBM calls the default in the affected versions. CISA added it to the Known Exploited Vulnerabilities catalog on 4 Aug 2026. ## What the flaw is Langflow is a tool for building AI agent and workflow applications. CVE-2026-9198 chains two flaws. First, an unauthenticated auto-login endpoint (/api/v1/auto_login) issues superuser bearer tokens to any caller on the network. Second, a code validation endpoint (/api/v1/validate/code) runs submitted Python when it defines a function: decorators, default arguments and annotations are executed at definition time. An attacker who can reach the server gets a superuser token from the first endpoint and then uses it to run code through the second. The result is remote code execution as the Langflow process, with no login and no user interaction. IBM says this affected default deployments where auto-login was enabled and the validation endpoint was reachable. IBM rates it 9.8 (CVSS 3.1 base score; network, low complexity, no privileges). The weakness is CWE-94, code injection. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | Langflow OSS runs a version from 1.0.0 through 1.10.0 | On the host run pip show langflow, or pip list \| grep -i langflow. In a container run docker exec CONTAINER pip show langflow. 1.10.1 is the first fixed release. | | auto-login is enabled (LANGFLOW_AUTO_LOGIN set to true), which IBM calls the default in the affected versions | Check the effective value: env \| grep LANGFLOW_AUTO_LOGIN, the .env file passed with --env-file, and for containers docker inspect --format '{{.Config.Env}}' CONTAINER. Langflow's current documentation says newer official packages and images ship with it set to false, and that an explicit true in the environment or .env file overrides that default. Treat an unset value on 1.10.0 or earlier as enabled. | | the server is reachable over the network by the attacker, so both API routes can be called | Run ss -ltnp \| grep -i langflow to see the bind address and port (Langflow documentation uses 7860 in its examples). Then check firewall or security group rules and any reverse proxy for that port from an untrusted network. Do not call the auto_login route on a live system to test it: on an affected version it issues a superuser token. | | no authenticating reverse proxy sits in front of the server | Review the reverse proxy configuration. Langflow's documentation says to deploy behind a reverse proxy with authentication and never to expose its ports directly to the internet. | ## Affected and fixed versions | Product line | Affected | First fixed | Notes | | --- | --- | --- | --- | | Langflow OSS (IBM bulletin; PyPI package langflow) | 1.0.0 through 1.10.0 | 1.10.1 (released 23 Jun 2026) | IBM bulletin published 2 Jul 2026; NVD configuration reads 1.0.0 up to but excluding 1.10.1 | The 1.10.1 release notes list two matching fixes without citing the CVE: 'do not execute code in validate_code' and 'stop issuing 365-day superuser token via auto_login'. ## If the fix cannot go in yet - Upgrade Langflow OSS to 1.10.1 or later. IBM strongly recommends this and lists no workarounds or mitigations. Cost: a change to authentication behavior, so read the next item before restarting an existing install. - Exposure-reducing step: set LANGFLOW_AUTO_LOGIN=false and set LANGFLOW_SUPERUSER_PASSWORD, as Langflow's documentation recommends. IBM limits the flaw to deployments with auto-login enabled, but does not list this as a workaround, so upgrade as well. Cost: users must sign in to the editor. On an existing auto-login install, prepare the account first (Langflow documentation): back up the database and configuration directory; keep the same database, configuration directory, encryption key and superuser username; set a known password on the existing account with PATCH /api/v1/users/{user_id} (the self-service reset endpoint needs the old password, so it does not suit an unknown generated one) and verify it with POST /api/v1/login; then set LANGFLOW_AUTO_LOGIN=false, LANGFLOW_SUPERUSER to that username and LANGFLOW_SUPERUSER_PASSWORD, and restart. Creating a different superuser does not transfer the flows. - Disabling auto-login does not revoke sessions or API keys issued earlier (Langflow documentation), and the 1.10.1 release notes name a 365-day superuser token. Plan to revoke or rotate credentials on any instance that was exposed. - Do not expose Langflow to untrusted networks and put an authenticating reverse proxy in front of it. Cost: clients must authenticate to the proxy. - CISA's KEV entry lists the required action: apply vendor mitigations, follow the forensic triage requirements it cites, and discontinue use of the product if mitigations are unavailable. ## What to look for - Search the reverse proxy or web server access log, and Langflow's own uvicorn access log, for the two routes. Langflow publishes no log lines for this CVE; these two are illustrations in the usual access-log format. A hit on the first route reads `203.0.113.7 - - [date] "GET /api/v1/auto_login HTTP/1.1" 200` and a hit on the second reads `"POST /api/v1/validate/code HTTP/1.1" 200`. A call to the first followed by the second from the same source, or from an address outside your network, is the sign to investigate. A 200 on auto_login from an outside address shows that a superuser token was issued. - On the host, review child processes spawned by the Langflow Python process, new files under temporary directories, and outbound connections to unfamiliar hosts. KEV marks forensic triage as required for this CVE, so preserve logs and memory before you clean up. - IBM, NVD and CISA publish no indicator addresses or file names for this CVE. Wiz Research's honeypot report lists these for attacks on its Langflow honeypot, without naming a CVE: staging server 94.26.106[.]29, a dropper at /tmp/x86_64 or /tmp/amd64 that deletes itself, and a miner binary renamed unicorn under /app/data/.claude/. Treat a match as a lead to investigate, not as proof of this CVE. ## Timeline | Date | Event | Source | | --- | --- | --- | | 23 Jun 2026 | Langflow 1.10.1 is released. Its notes list 'do not execute code in validate_code' and 'stop issuing 365-day superuser token via auto_login'. | [source](https://github.com/langflow-ai/langflow/releases/tag/v1.10.1) | | 2 Jul 2026 | IBM publishes the security bulletin (initial publication) and credits Harish Kolla. | [source](https://www.ibm.com/support/pages/node/7278927) | | 17 Jul 2026 | NVD publishes CVE-2026-9198. The CISA coordinator SSVC entry on the record carries the same date, with exploitation active, automatable yes, technical impact total. This is CISA's assessment stamp and the earliest dated mark of exploitation, not an observed attack. No report names the attacker or the first attack. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-9198) | | 4 Aug 2026 | CISA adds the CVE to KEV with forensic triage required and a due date of 7 Aug 2026. Ransomware use is listed as Unknown. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-9198) | The earliest dated mark of exploitation is CISA's, on 17 Jul 2026, 15 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far Wiz Research's report of 27 Aug 2026 covers 90 days of telemetry from honeypots that include Langflow and records attackers staging tools on the Langflow honeypot. It names no CVE and gives no date for those attacks, so it does not date the first exploitation of CVE-2026-9198. ## Why runtime reachability settles it The default configuration of an affected version is vulnerable, so version alone does not decide exposure: the running instance must have auto-login on and be reachable by an outsider. Those are runtime facts, shown by the effective LANGFLOW_AUTO_LOGIN value and the listening address. KEV and CISA's SSVC entry confirm active attacks. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-53266: Linux kernel](/ledger/cve-2026-53266/) - [CVE-2026-59822: BerriAI LiteLLM, LiteLLM MCP endpoint auth bypass](/ledger/cve-2026-59822/) - [CVE-2026-46300: Linux Kernel](/ledger/cve-2026-46300/) ## Sources - CVE-2026-9198: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-9198 (read 19 Nov 2026) - CVE-2026-9198: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-9198 (read 19 Nov 2026) - CVE-2026-9198: vendor advisory (IBM security bulletin): https://www.ibm.com/support/pages/node/7278927 (read 19 Nov 2026) - CVE-2026-9198: vendor release notes for 1.10.1: https://github.com/langflow-ai/langflow/releases/tag/v1.10.1 (read 19 Nov 2026) - CVE-2026-9198: vendor documentation on authentication and LANGFLOW_AUTO_LOGIN: https://docs.langflow.org/api-keys-and-authentication (read 19 Nov 2026) - CVE-2026-9198: GitHub advisory database record: https://api.github.com/advisories?cve_id=CVE-2026-9198 (read 19 Nov 2026) - CVE-2026-9198: Wiz Research honeypot report: https://www.wiz.io/blog/ai-infrastructure-honeypot (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-9198/ --- # CVE-2026-53266: Linux kernel, is it exploitable where you run it Ledger entry, disclosed 1 Jun 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-53266 affects Linux kernel and was disclosed 1 Jun 2026 with a CVSS score of 8.8. Exposure needs these conditions: version: the kernel is in an affected range and carries no fix or distribution backport; an ebtables nat rule uses the snat target with ARP rewrite, so the ebtable_nat and ebt_snat modules are loaded. ## What the flaw is ebtables is the Linux tool for filtering and rewriting Ethernet frames on a bridge. Its snat target changes the source MAC address. With the --snat-arp option it also rewrites the sender hardware address inside ARP packets. The ARP rewrite used skb_store_bits() to write directly into the packet buffer, but it never made that range writable first. If the range sat in a nonlinear buffer fragment that is backed by a file page imported through splice, the kernel wrote the new MAC address straight into the page cache copy of that file. The fix calls skb_ensure_writable() on the ARP range before the header is read and before the write. CISA describes the result as an out-of-bounds write into a file-backed page. Red Hat says a local attacker can reach privilege escalation, memory corruption or denial of service, and that the impact is limited to hosts with the special bridge netfilter rules. The bytes written are a MAC address. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | Version: the kernel is in an affected range and carries no fix or distribution backport. | Run `uname -r` and compare with the versions table. On RPM systems run `rpm -q --changelog kernel \| grep CVE-2026-53266`. On Debian and Ubuntu run `apt changelog linux-image-$(uname -r) \| grep CVE-2026-53266`, or use the Debian and Ubuntu trackers. | | An ebtables nat rule uses the snat target with ARP rewrite, so the ebtable_nat and ebt_snat modules are loaded. | Run `ebtables -t nat -L` as root and look for `-j snat` rules that include `--snat-arp` (the ebtables manual describes this option as changing the hardware source address inside ARP messages). Also check rule files that load at boot, for example /etc/sysconfig/ebtables or your firewall tooling. Then run `lsmod \| grep -E 'ebtable_nat\|ebt_snat'`. If neither module is loaded and no ebtables nat rule exists, the vulnerable code does not run, but loading state changes when a rule is added. | | A bridge in the same network namespace carries the ARP traffic. | Run `ip link show type bridge` and `bridge link`. Repeat inside container and VM host network namespaces that run their own bridges, for example with `ip netns exec ebtables -t nat -L`. | | An attacker can run code locally as a low-privilege user and can reach CAP_NET_ADMIN over the namespace that holds the rule. | Run `sysctl user.max_user_namespaces` (0 means no new user namespaces) and list local accounts that can run code. The kernel source requires CAP_NET_ADMIN in the owning user namespace to change ebtables rules. | | The kernel line still gets fixes. | Kernel 5.4 ended with 5.4.302 on 3 Dec 2025 and has no fix. On such hosts only a move to a supported kernel or the Red Hat mitigation removes the exposure. | ## Affected and fixed versions | Product or line | Affected | First fixed | | --- | --- | --- | | Linux 5.4 | 5.4.73 to 5.4.302 (last release) | No fix; the line is end-of-life | | Linux 5.8 | 5.8.17 and later 5.8 releases | No stable fix; the line is end-of-life | | Linux 5.9 and 5.10 | 5.9.2 and later 5.9 releases, and 5.10.0 to before 5.10.259 | 5.10.259 (19 Jun 2026). The 5.9 line is end-of-life with no fix | | Linux 5.15 | 5.11 to before 5.15.210 | 5.15.210 (19 Jun 2026) | | Linux 6.1 | 5.16 to before 6.1.176 | 6.1.176 (19 Jun 2026) | | Linux 6.6 | 6.2 to before 6.6.143 | 6.6.143 (19 Jun 2026) | | Linux 6.12 | 6.7 to before 6.12.94 | 6.12.94 (19 Jun 2026) | | Linux 6.18 | 6.13 to before 6.18.36 | 6.18.36 (19 Jun 2026) | | Linux 7.0 | 6.19 to before 7.0.13 | 7.0.13 (19 Jun 2026) | | Linux 7.1 | 7.1-rc1 to 7.1-rc6 | 7.1 (14 Jun 2026), mainline commit 67ba971ae025 in 7.1-rc7 | | RHEL 10 / 10.0 EUS / NVIDIA for RHEL 10 | All earlier kernels | kernel-6.12.0-211.59.1.el10_2 (RHSA-2026:71233, 24 Sep 2026) / kernel-6.12.0-55.106.1.el10_0 (RHSA-2026:71326) / kernel-6.12.0-231.20.el10nv (RHSA-2026:72059) | | RHEL 9 / 9.6 EUS / 9.4 / 9.2 extended | All earlier kernels | kernel-5.14.0-687.23.1.el9_8 (RHSA-2026:36645, 8 Jul 2026) / 5.14.0-570.143.1.el9_6 (RHSA-2026:71327) / 5.14.0-427.152.1.el9_4 (RHSA-2026:71569) / 5.14.0-284.194.1.el9_2 (RHSA-2026:71601) | | RHEL 8 and extended streams | All earlier kernels | kernel-4.18.0-553.143.1.el8_10 (RHSA-2026:39083, 14 Jul 2026) / 4.18.0-477.169.1.el8_8 (RHSA-2026:71325) / 4.18.0-372.217.1.el8_6 (RHSA-2026:70797) / 4.18.0-305.209.1.el8_4 (RHSA-2026:71565) | | RHEL 7 ELS / RHEL 6 ELS | All earlier kernels | kernel-3.10.0-1160.164.1.el7 (RHSA-2026:71687, 24 Sep 2026) / kernel-2.6.32-754.64.1.el6 (RHSA-2026:73732, 30 Sep 2026) | | Debian | Earlier source packages | bullseye 5.10.259-1 (DLA-4664-1), bookworm 6.1.176-1 (DLA-4665-1), trixie 6.12.94-1, unstable 7.0.13-1; linux-6.1 on bullseye 6.1.176-1~deb11u1 (DLA-4671-1) | | Ubuntu linux package | Earlier kernels | jammy 5.15.0-198.208, noble 6.8.0-146.146, resolute 7.0.0-31.31, jammy HWE 6.8.0-149.149~22.04.1. Focal has no fix yet (needed); questing is end-of-life. Trusty, xenial and bionic are not affected | NVD's lower bounds (5.4.73, 5.8.17, 5.9.2) are the stable releases that first picked up commit 63137bc5882a, which the fix cites as its origin. Releases before those do not carry the bug. Red Hat kpatch errata of 28 Oct 2026 exist for RHEL 8 (RHSA-2026:92062), RHEL 9 (RHSA-2026:91650) and RHEL 10 (RHSA-2026:92061). ## If the fix cannot go in yet - Install a kernel with the fix and reboot. Cost: a reboot. Red Hat also ships kpatch updates for some streams. - Red Hat mitigation: disable ARP hardware address rewriting in ebtables SNAT rules, or remove ebtables SNAT rules that act on ARP traffic on bridge interfaces. Cost: the bridge no longer rewrites the ARP sender address, so any set-up that depends on that rewrite (for example MAC address translation for bridged virtual machines or containers) stops working. - CISA's KEV description warns the product could be end-of-life and advises moving to a supported version. Cost: a kernel upgrade project. - Optional hardening beyond the vendor text: unload ebtable_nat and ebt_snat if no rule needs them. Cost: all ebtables nat rules stop working. ## What to look for - CISA's KEV entry marks forensic triage as required and gives no indicators. Red Hat, Debian, Ubuntu and the kernel project publish no log strings, file paths or addresses for this CVE. - The write goes to a cached file page, so a checksum of the file on disk does not show it. Review local accounts that could create network namespaces and ebtables rules on hosts that run bridges. ## Timeline | Date | Event | Source | | --- | --- | --- | | 14 Oct 2020 | Commit 63137bc5882a (ebtables: fix dropping of small packets in bridge nat) is written. The 2026 fix cites it as the origin. Stable 5.4.73, 5.8.17 and 5.9.2 carry it. | [source](https://api.github.com/repos/torvalds/linux/commits/63137bc5882a) | | 23 May 2026 | The fix is written; Yiming Qian is credited as reporter. | [source](https://api.github.com/repos/torvalds/linux/commits/67ba971ae02514d85818fe0c32549ab4bfa3bf49) | | 1 Jun 2026 | The fix is committed to mainline (7.1-rc7). Red Hat sets its public date to the same day and rates the flaw Important. | [source](https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-53266.json) | | 14 Jun 2026 | Linux 7.1 is released with the fix. | [source](https://cdn.kernel.org/pub/linux/kernel/v7.x/ChangeLog-7.1) | | 19 Jun 2026 | Stable 5.10.259, 5.15.210, 6.1.176, 6.6.143, 6.12.94, 6.18.36 and 7.0.13 ship the backports. | [source](https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.1.176) | | 25 Jun 2026 | The kernel CNA publishes CVE-2026-53266 and NVD lists it (CVSS 8.8). | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-53266) | | 8 Jul 2026 | Red Hat ships the RHEL 9 errata RHSA-2026:36645; RHEL 8 follows on 14 Jul 2026 (RHSA-2026:39083). | [source](https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-53266.json) | | 18 Sep 2026 | CISA adds CVE-2026-53266 to the Known Exploited Vulnerabilities catalog, due 21 Sep 2026, with the note that the product could be end-of-life. | [source](https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json) | | 24 Sep 2026 | Red Hat ships RHEL 10 and RHEL 7 ELS errata; RHEL 6 ELS follows on 30 Sep 2026 and kpatch errata on 28 Oct 2026. | [source](https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-53266.json) | The earliest dated mark of exploitation is CISA's, on 18 Sep 2026, 109 days after disclosure; exploitation may have begun earlier. ## Exploitation reported so far CISA added the flaw to KEV on 18 Sep 2026 and lists ransomware use as unknown. The KEV entry gives no details of the attacks. The CISA entry in the NVD record carries an SSVC decision of exploitation active, automatable no and technical impact total, stamped 18 Sep 2026. Neither CISA, Red Hat, Debian, Ubuntu nor the kernel project names the actor or the targets, and no researcher has published a description of the attacks. ## Why runtime reachability settles it The flaw only runs when an ebtables nat rule with ARP rewrite exists on a bridge and a local user can reach it. The kernel version alone cannot show that, but reading the loaded modules and the ebtables rules on the running host can: that is runtime state, not source code. Known-exploited status says to act, and the rule check says where. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-46300: Linux Kernel](/ledger/cve-2026-46300/) - [CVE-2026-9198: IBM Langflow OSS](/ledger/cve-2026-9198/) - [CVE-2026-34486: Apache Software Foundation Apache Tomcat](/ledger/cve-2026-34486/) ## Sources - CVE-2026-53266: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-53266 (read 19 Nov 2026) - CVE-2026-53266: CISA KEV feed (catalog version 2026.11.18): https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - CVE-2026-53266: vendor advisory data (Red Hat Security Data API: errata, package versions, mitigation, statement): https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-53266.json (read 19 Nov 2026) - CVE-2026-53266: upstream fix commit (GitHub mirror of the Linux kernel); the seven stable commits in the KEV notes were read the same way: https://api.github.com/repos/torvalds/linux/commits/67ba971ae02514d85818fe0c32549ab4bfa3bf49 (read 19 Nov 2026) - CVE-2026-53266: upstream commit named as the origin of the bug: https://api.github.com/repos/torvalds/linux/commits/63137bc5882a (read 19 Nov 2026) - CVE-2026-53266: kernel.org stable change log 6.1.176 (also read for 5.10.259, 5.15.210, 6.6.143, 6.12.94, 6.18.36, 7.0.13, 7.1, 5.4.73, 5.8.17, 5.9.2 and 5.4.300 to 5.4.302): https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.1.176 (read 19 Nov 2026) - CVE-2026-53266: Linux 6.12 kernel source, ebtables.c (capability check for rule changes): https://raw.githubusercontent.com/torvalds/linux/v6.12/net/bridge/netfilter/ebtables.c (read 19 Nov 2026) - CVE-2026-53266: Linux 6.12 kernel source, ebt_snat.c (code before the fix): https://raw.githubusercontent.com/torvalds/linux/v6.12/net/bridge/netfilter/ebt_snat.c (read 19 Nov 2026) - CVE-2026-53266: ebtables manual page (snat target and --snat-arp option): https://ebtables.netfilter.org/misc/ebtables-man.html (read 19 Nov 2026) - CVE-2026-53266: Debian security tracker: https://security-tracker.debian.org/tracker/CVE-2026-53266 (read 19 Nov 2026) - CVE-2026-53266: Ubuntu CVE tracker data: https://ubuntu.com/security/cves/CVE-2026-53266.json (read 19 Nov 2026) - CVE-2026-53266: GitHub advisory database entry GHSA-xmhg-rr34-2chj: https://api.github.com/advisories?cve_id=CVE-2026-53266 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-53266/ --- # CVE-2026-46300: Linux Kernel, is it exploitable where you run it Ledger entry, disclosed 13 May 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-46300 affects Linux Kernel and was disclosed 13 May 2026 with a CVSS score of 7.8. Exposure needs these conditions: the running kernel lacks the fix or a distribution backport; a local user can run code, and esp4 or esp6 can load with unprivileged namespaces allowed. CISA added it to the Known Exploited Vulnerabilities catalog on 28 Oct 2026. ## What the flaw is Linux keeps file contents in memory in a page cache. Network code marks packet buffers that point into those pages as shared, so that nobody modifies them in place. The function skb_try_coalesce() merges receive buffers and, before the fix, dropped that marker when it moved the page fragments into the merged buffer. The IPsec ESP input code trusts the marker. When it sees none, it decrypts in place and skips the copy it would otherwise make. The decrypted bytes then land in the page cache of a file that was only read, for example a setuid binary. The file on disk is unchanged, but the next program run from the cache uses the altered copy. A low-privilege local user can use this to gain root. A public proof of concept named Fragnesia appeared on 13 May 2026 and belongs to the Dirty Frag class of page-cache bugs. This page gives no exploit steps. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | the running kernel lacks the fix or a distribution backport | Run `uname -r`. Compare it with the versions table; kernels older than 5.6 (Ubuntu 18.04 and 20.04 kernels) lack espintcp, which the public proofs of concept need, so they cannot run them (Ubuntu). On RPM systems run `rpm -q --changelog kernel \| grep CVE-2026-46300`. On Debian or Ubuntu run `apt changelog linux-image-$(uname -r) \| grep CVE-2026-46300` or look up the package in the Debian and Ubuntu trackers. | | a local user can run code, and esp4 or esp6 can load with unprivileged namespaces allowed | Foothold: list who can log in or run code with `getent passwd`, `last`, web application accounts, CI runners and container users; without a local foothold the flaw is not reachable (CVSS AV:L, PR:L). ESP modules: Red Hat documents `lsmod \| grep -E 'esp4\|esp6'`; no output means IPsec is not in active use, but it does not show that the modules cannot load later. Run `modinfo esp4` to see whether the module exists for the running kernel, and `cat /etc/modprobe.d/dirtyfrag.conf` to see whether the Red Hat block is in place. Namespaces: `sysctl user.max_user_namespaces`; 0 blocks creation of new user namespaces (Red Hat). Red Hat documents the module block and the namespace setting as separate mitigations, and either one stops the path, so a host is exposed only while both are open. | ## Affected and fixed versions | Product or line | Affected | First fixed | | --- | --- | --- | | Linux 3.9 to 5.10 | 3.9 through 5.10.257 (NVD) | No fixed upstream 5.10 stable release through 5.10.274. Debian bullseye package 5.10.257-1 (DLA-4606-1) is fixed (see note) | | Linux 5.11 to 5.15 | 5.11 before 5.15.208 | 5.15.208 | | Linux 5.16 to 6.1 | 5.16 before 6.1.174 | 6.1.174 | | Linux 6.2 to 6.6 | 6.2 before 6.6.141 | 6.6.141 | | Linux 6.7 to 6.12 | 6.7 before 6.12.91 | 6.12.91 | | Linux 6.13 to 6.18 | 6.13 before 6.18.33 | 6.18.33 | | Linux 6.19 to 7.0 | 6.19 before 7.0.10 | 7.0.10 | | Linux 7.1 candidates | 7.1-rc1 to 7.1-rc4 | 7.1-rc5 (mainline) | | RHEL 10 / 10.0 EUS | All earlier kernels | kernel-6.12.0-211.16.1.el10_2 (RHSA-2026:19569, 20 May 2026) / kernel-6.12.0-55.75.1.el10_0 (RHSA-2026:20299) | | RHEL 9 / 9.6 EUS / 9.4 EUS | All earlier kernels | kernel-5.14.0-687.10.1.el9_8 (RHSA-2026:19568) / 5.14.0-570.116.1.el9_6 (RHSA-2026:20129) / 5.14.0-427.126.1.el9_4 (RHSA-2026:20054) | | RHEL 8 / 8.8 / 8.6 / 8.4 extended support | All earlier kernels | kernel-4.18.0-553.125.1.el8_10 (RHSA-2026:19666) / 4.18.0-477.143.1.el8_8 (RHSA-2026:19521) / 4.18.0-372.193.1.el8_6 (RHSA-2026:20051) / 4.18.0-305.192.1.el8_4 (RHSA-2026:20130) | | RHEL 6 and 7 | Not affected (Red Hat) | Not applicable | | OpenShift Container Platform 4.12 to 4.21 | RHCOS images before the errata | One RHSA per minor version, for example 4.18: RHSA-2026:21656 (3 Jun 2026) and 4.21: RHSA-2026:23240 (11 Jun 2026) | | Debian | Earlier source packages | bullseye 5.10.257-1 (DLA-4606-1), bookworm 6.1.174-1 (DSA-6306-1), trixie 6.12.90-1 (DSA-6295-1), unstable 7.0.9-1; later builds in the tracker on 19 Nov 2026: bookworm (security) 6.1.192-2, trixie (security) 6.12.117-1 | | Ubuntu linux package | Earlier kernels | focal 5.4.0-233.253, jammy 5.15.0-181.191, noble 6.8.0-124.124, questing 6.17.0-35.35, resolute 7.0.0-22.22 | | Siemens SIMATIC S7-1500 CPU 1518(F)-4 PN/DP MFP and SIPLUS variants (a separate product, not a Linux distribution: read the CPU firmware version in TIA Portal online and diagnostics or on the CPU web server) | V3.1.5 (no fix) and V3.1.6 | V3.1.7 | NVD lists 5.10.257 as the last affected upstream 5.10 release, and the kernel.org change logs for 5.10.258 through 5.10.274 do not contain the fix, so users of upstream 5.10 depend on the distribution. Debian's tracker lists bullseye 5.10.257-1 (DLA-4606-1) and trixie 6.12.90-1 (DSA-6295-1) as fixed. Those are Debian package versions: Debian ships the fix as its own patch on the upstream release of that number, so the package is fixed even though upstream 5.10.257 is the last affected release and upstream 6.12.91 is the first fixed 6.12 release. Compare Debian and other distribution kernels with the distribution's package version, not the upstream number, because distributions backport the fix without changing the upstream version. ## If the fix cannot go in yet - Install the fixed kernel from your distribution and reboot. Cost: a reboot. Red Hat and Siemens also ship the fix, and Red Hat offers kpatch for some streams, which avoids the reboot. - Block the IPsec ESP modules (Red Hat's broader method). Write install lines for esp4 and esp6 that run /bin/false into /etc/modprobe.d/dirtyfrag.conf and run rmmod esp4 esp6. Cost: IPsec stops working after the next reboot. Revert by deleting the file. Check first with lsmod | grep -E 'esp4|esp6': if neither module is loaded, IPsec is not in use and blocking is safe. - Disable unprivileged user namespaces (Red Hat's targeted method when IPsec must stay). Set user.max_user_namespaces=0 in /etc/sysctl.d/dirtyfrag.conf and run sysctl --system. Cost: rootless containers, sandboxed browsers and Flatpak can stop working. - Limit local access (Red Hat). Turn off unneeded SSH access, keep SELinux in enforcing mode, use the default OpenShift Security Context Constraints, run workloads as non-root and restrict oc debug to trusted administrators. Red Hat notes this list is not exhaustive and closing one access path does not close the others. - Keep the module mitigation in place after patching for now. The researcher who posted the later Dirty Frag fixes advised on 22 May 2026 that more variants in the ESP in-place path are expected. - On the Siemens CPU 1518(F)-4 PN/DP MFP, update to V3.1.7. Siemens also advises building and running only applications from trusted sources and limiting access to the interactive shell of the additional GNU/Linux subsystem to trusted staff. ## What to look for - CISA's KEV entry says the flaw was chained with CVE-2026-109618 and CVE-2026-109619 to implant PHP webshells and run commands as root, and marks forensic triage as required. CISA publishes no file names, hashes or addresses in the entry. - Reads go through the page cache, so a package verification run before a reboot or cache drop can show the change: run `rpm -V PACKAGE` (RPM systems) or `debsums -s PACKAGE` (Debian and Ubuntu) for the package that owns each setuid binary, for example `rpm -qf /usr/bin/su`, and compare a `sha256sum` of the running binary with the one in a clean copy of the package. A mismatch means the cached copy was altered. Dropping the cache with `echo 3 > /proc/sys/vm/drop_caches` or rebooting restores the on-disk bytes and erases that evidence, so record the hashes first. Red Hat, Ubuntu and Siemens publish no indicators of compromise for this CVE. - Review unexpected PHP files in web roots and new root-level processes on hosts where an untrusted local user or web application account could have run code. ## Timeline | Date | Event | Source | | --- | --- | --- | | 13 May 2026 | Fragnesia is disclosed on oss-security by the V12 team (discoverer William Bowling) with a public proof of concept. The fix patch, authored the same day, is already posted to netdev, and Greg Kroah-Hartman assigns CVE-2026-46300 to it. Red Hat gives the same public date and rates the flaw Important. | [source](http://www.openwall.com/lists/oss-security/2026/05/13/5) | | 14 May 2026 | CISA's SSVC entry on the NVD record carries the timestamp 14 May 2026 and marks exploitation: active, automatable: no, technical impact: total. | [source](https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-46300) | | 15 May 2026 | Jakub Kicinski commits the standalone fix f84eca5817390257cef78013d0112481c503b4a3 to the netdev tree, two days after the patch was authored and the CVE assigned. | [source](https://api.github.com/repos/torvalds/linux/commits/f84eca5817390257cef78013d0112481c503b4a3) | | 20 May 2026 | Red Hat releases the first errata for RHEL 8, 9 and 10 (RHSA-2026:19666, RHSA-2026:19568, RHSA-2026:19569). | [source](https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-46300.json) | | 21 May 2026 | The patch author for the broader Dirty Frag variants reports on oss-security that fix 48f6a5356a33 is merged into netdev and that the CVE-2026-46300 patch was split out. He advises keeping the module mitigation in place. | [source](http://www.openwall.com/lists/oss-security/2026/05/21/11) | | 23 May 2026 | Stable kernels 5.15.208, 6.1.174, 6.6.141, 6.12.91, 6.18.33 and 7.0.10 ship the backport. NVD publishes the record the same day. | [source](https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.1.174) | | 3 Jul 2026 | Red Hat updates bulletin RHSB-2026-003 (Dirty Frag and Fragnesia), status Resolved, with the module and namespace mitigations. | [source](https://access.redhat.com/security/vulnerabilities/RHSB-2026-003) | | 14 Jul 2026 | Siemens publishes SSA-019113 for SIMATIC S7-1500 CPU 1518(F)-4 PN/DP MFP firmware V3.1.6. Version 1.1 of 8 Sep 2026 adds the fix and version 1.2 of 13 Oct 2026 links V3.1.7. | [source](https://cert-portal.siemens.com/productcert/html/ssa-019113.html) | | 28 Oct 2026 | CISA adds CVE-2026-46300 to the Known Exploited Vulnerabilities catalog with a due date of 31 Oct 2026. | [source](https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json) | The earliest dated mark of exploitation is CISA's, on 14 May 2026, 1 day after disclosure; exploitation may have begun earlier. ## Exploitation reported so far The 14 May 2026 date is CISA's own assessment date on the SSVC entry, which NVD published nine days later, on 23 May 2026. The entry gives no source for the finding, so it is not a dated incident report. CISA's first public statement of use is the KEV listing. The KEV entry says the flaw can be chained with CVE-2026-109618 and CVE-2026-109619 to implant PHP webshells and run commands as root, and lists ransomware use as unknown. A working public proof of concept has existed since 13 May 2026. Neither CISA nor Red Hat, Ubuntu or Siemens publishes the actor, victims or indicators. ## Why runtime reachability settles it Exposure needs a local foothold plus a kernel that can load the ESP code and lets the user create namespaces. Those are facts about the running host, so checking the kernel, the loaded modules and the namespace setting settles it, and a severity score or the version alone cannot. Known-exploited status tells you to act now, but not which hosts are reachable. See [runtime reachability](/methods/runtime-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-34486: Apache Software Foundation Apache Tomcat](/ledger/cve-2026-34486/) - [CVE-2026-53266: Linux kernel](/ledger/cve-2026-53266/) - [CVE-2026-123158: OpenBSD OpenSSH, OpenSSH musl no-PAM password bypass](/ledger/cve-2026-123158/) ## Sources - CVE-2026-46300: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-46300 (read 19 Nov 2026) - CVE-2026-46300: CISA KEV feed (catalog version 2026.11.18): https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - CVE-2026-46300: vendor advisory (Red Hat CVE page): https://access.redhat.com/security/cve/CVE-2026-46300 (read 19 Nov 2026) - CVE-2026-46300: vendor advisory data (Red Hat Security Data API: errata and package versions): https://access.redhat.com/hydra/rest/securitydata/cve/CVE-2026-46300.json (read 19 Nov 2026) - CVE-2026-46300: vendor advisory (Red Hat security bulletin RHSB-2026-003, mitigation): https://access.redhat.com/security/vulnerabilities/RHSB-2026-003 (read 19 Nov 2026) - CVE-2026-46300: vendor advisory (Siemens SSA-019113, firmware V3.1.6): https://cert-portal.siemens.com/productcert/html/ssa-019113.html (read 19 Nov 2026) - CVE-2026-46300: vendor advisory (Siemens SSA-082556, firmware V3.1.5): https://cert-portal.siemens.com/productcert/html/ssa-082556.html (read 19 Nov 2026) - CVE-2026-46300: oss-security disclosure and CVE assignment: http://www.openwall.com/lists/oss-security/2026/05/13/5 (read 19 Nov 2026) - CVE-2026-46300: oss-security post on the Dirty Frag variant fixes: http://www.openwall.com/lists/oss-security/2026/05/21/11 (read 19 Nov 2026) - CVE-2026-46300: oss-security reply in the same thread: http://www.openwall.com/lists/oss-security/2026/05/21/12 (read 19 Nov 2026) - CVE-2026-46300: oss-security reply in the same thread: http://www.openwall.com/lists/oss-security/2026/05/21/13 (read 19 Nov 2026) - CVE-2026-46300: upstream fix commit (GitHub mirror of the Linux kernel): https://api.github.com/repos/torvalds/linux/commits/f84eca5817390257cef78013d0112481c503b4a3 (read 19 Nov 2026) - CVE-2026-46300: kernel.org stable change log 6.1.174 (also read for 5.15.208, 6.6.141, 6.12.91, 6.18.33, 7.0.10 and 5.10.258 to 5.10.274): https://cdn.kernel.org/pub/linux/kernel/v6.x/ChangeLog-6.1.174 (read 19 Nov 2026) - CVE-2026-46300: Debian security tracker: https://security-tracker.debian.org/tracker/CVE-2026-46300 (read 19 Nov 2026) - CVE-2026-46300: Ubuntu CVE tracker data: https://ubuntu.com/security/cves/CVE-2026-46300.json (read 19 Nov 2026) - CVE-2026-46300: GitHub advisory database entry GHSA-47jg-vqrv-5f8v: https://api.github.com/advisories?cve_id=CVE-2026-46300 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-46300/ --- # CVE-2026-34486: Apache Software Foundation Apache Tomcat, exploitable where you run it Ledger entry, disclosed 9 Apr 2026, sources read 19 Nov 2026 **The short answer.** CVE-2026-34486 affects Apache Software Foundation Apache Tomcat and was disclosed 9 Apr 2026 with a CVSS score of 7.5. Exposure needs these conditions: version: the node runs Tomcat 11.0.20, 10.1.53 or 9.0.116; the Cluster channel in server.xml uses the EncryptInterceptor, and its receiver port (default 4000) is reachable from hosts outside the cluster. ## What the flaw is Tomcat can replicate user sessions between nodes over a cluster channel. The EncryptInterceptor is the component that encrypts those channel messages. CVE-2026-29146 covered a padding oracle weakness in its default cipher mode (CBC). The fix for it shipped in Tomcat 11.0.20, 10.1.53 and 9.0.116, and that fix contained an error, and CVE-2026-34486 is that error: it allows the EncryptInterceptor to be bypassed, so data on the channel is not protected as intended (CWE-311, missing encryption of sensitive data). Apache's own severity rating for it is Important; the CVSS base score is separate (next paragraph). The CVSS 3.1 base score is 7.5, with high impact on confidentiality only. The practical risk is larger than that CVSS score suggests. SOCRadar reports that attackers used a public Java deserialization toolkit against the cluster port of Tomcat 9.0.116 servers and ran commands on them, which means that a bypassed channel can carry attacker-supplied serialized data into the node. CISA's KEV entry adds that the flaw can be chained with CVE-2025-24813, a Tomcat flaw in partial PUT handling of the default servlet that can allow code execution when its own conditions hold. Only three exact releases are affected, so most Tomcat servers are not exposed to this CVE. Servers that do not use clustering, or use it without the EncryptInterceptor, are outside the flaw. ## Are you exposed: each condition and how to check it | Condition for exposure | How to check it on your system | | --- | --- | | Version: the node runs Tomcat 11.0.20, 10.1.53 or 9.0.116. | Run version.sh (Linux) or version.bat (Windows) in the Tomcat bin directory, or catalina.sh version. Read the Server number line. On a package install run rpm -q tomcat or dpkg -l \| grep tomcat and compare with the distribution errata, since backported builds keep older version numbers. | | The Cluster channel in server.xml uses the EncryptInterceptor, and its receiver port (default 4000) is reachable from hosts outside the cluster. | grep -n -e 'Cluster' -e 'EncryptInterceptor' $CATALINA_BASE/conf/server.xml. Both must appear outside XML comments, with the Interceptor element (class org.apache.catalina.tribes.group.interceptors.EncryptInterceptor) inside the Channel element. Then find the Receiver element (class org.apache.catalina.tribes.transport.nio.NioReceiver, default port 4000, autoBind 100, so any port from 4000 to 4100). On the host run ss -ltnp \| grep java and look for those ports. From a host outside the cluster run nc -vz 4000, and review iptables -S, nft list ruleset or firewall-cmd --list-all and the network ACL or security group for rules on the port range. The condition holds if the outside host connects. | | Only for the chain CISA mentions: the conditions of CVE-2025-24813 also hold. These are, per the Tomcat security page, writes enabled for the default servlet (off by default), partial PUT support (on by default), file-based session persistence at its default location, and a library on the class path that can be abused through deserialization. | In conf/web.xml look at the default servlet init-param readonly (a value of false enables writes). In conf/context.xml or the web app, look for a PersistentManager with FileStore. Check the Tomcat version against the CVE-2025-24813 fix (11.0.3, 10.1.35 and 9.0.99, per the Tomcat security pages). | ## Affected and fixed versions | Tomcat branch | Affected version | First fixed version | Fix released (per Tomcat changelog) | | --- | --- | --- | --- | | 11.0.x | 11.0.20 only | 11.0.21 | 4 Apr 2026 | | 10.1.x | 10.1.53 only | 10.1.54 | 2 Apr 2026 | | 9.0.x | 9.0.116 only | 9.0.117 | 3 Apr 2026 | Apache lists no other release as affected. Earlier releases of each branch carry the older CVE-2026-29146 flaw instead, which Apache lists as a separate issue. Apache announced the CVE on 9 Apr 2026, after all three fixed releases were out. Distributions such as RHEL backport fixes under their own package versions, so check the distribution errata for packaged installs. ## If the fix cannot go in yet - Upgrade to Tomcat 11.0.21, 10.1.54 or 9.0.117 or later. This is the vendor fix. It needs a restart of each node, so plan a rolling restart if sessions are replicated. Apache lists no workaround. - Limit the cluster receiver port to the other cluster nodes with a host firewall or network ACL. The Tomcat clustering guide says clusters are meant for a trusted network. Cost: any host that is not a cluster member loses access to that port, and a node added to the cluster later needs a new rule. - Where a trusted network is not available, the Tomcat 11 clustering guide describes a TLS-protected channel (secure attribute on the Channel, pre-shared key and identity, securePort on the Receiver). Cost: it needs OpenSSL-FFM on Java 22 or later, or Tomcat Native 1.2.17 or later, and the same settings on every node. The Tomcat 11 configuration reference also marks the EncryptInterceptor as deprecated and says it should no longer be used to secure cluster traffic. - Red Hat states that no mitigation meeting its criteria exists for its Tomcat packages, so RHEL and JBoss Web Server users take the vendor errata. ## What to look for - SOCRadar indicators for the campaign that used this CVE: attacker hosts 130[.]94[.]17[.]180 and 130[.]94[.]30[.]168 (payload served from port 8080 path /slt), secondary host 92[.]38[.]135[.]196 on port 8888, and 149[.]129[.]37[.]105 on port 30005. Also look for outbound curl or wget from the Tomcat process piped to sh, a file at /tmp/log_de.log, a process that names itself [kworker/0:2], and memfd_create with fexecve in EDR telemetry. SOCRadar also lists /error.jsp and /tunnel.jsp web shells, but those belong to the CVE-2025-24813 part of the campaign. - Unexpected child processes of the Tomcat Java process (shells, curl, wget) and connections to the cluster port from hosts outside the cluster are the generic signs of exploitation of this class of flaw. ## Timeline | Date | Event | Source | | --- | --- | --- | | 20 Mar 2026 | Tomcat 11.0.20 is released with the flawed fix for CVE-2026-29146. | [source](https://tomcat.apache.org/tomcat-11.0-doc/changelog.html) | | 26 Mar 2026 | The error in the fix is reported to the Tomcat security team (credit: Bartlomiej Dmitruk, striga.ai). | [source](https://tomcat.apache.org/security-11.html) | | 2 Apr 2026 | Tomcat 10.1.54 is released with the fix. | [source](https://tomcat.apache.org/tomcat-10.1-doc/changelog.html) | | 3 Apr 2026 | Tomcat 9.0.117 is released with the fix. | [source](https://tomcat.apache.org/tomcat-9.0-doc/changelog.html) | | 4 Apr 2026 | Tomcat 11.0.21 is released with the fix. | [source](https://tomcat.apache.org/tomcat-11.0-doc/changelog.html) | | 9 Apr 2026 | Apache publishes the advisory and NVD publishes the record. The advisory has no later revision. | [source](https://lists.apache.org/thread/9510k5p5zdvt9pkkgtyp85mvwxo2qrly) | | 24 Apr 2026 | Start of the range (about 24 to 29 Apr) in which SOCRadar's timeline places exploitation of this CVE against Taiwanese Tomcat servers. | [source](https://socradar.io/blog/snowlight-government-chinese-campaign/) | | 31 Jul 2026 | SOCRadar publishes its report on the SNOWLIGHT campaign, naming this CVE. | [source](https://socradar.io/blog/snowlight-government-chinese-campaign/) | | 4 Aug 2026 | CISA adds the CVE to KEV with a due date of 7 Aug 2026 and the note that it can be chained with CVE-2025-24813. | [source](https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-34486) | Exploited 15 days after disclosure: disclosed 9 Apr 2026, first exploitation reported 24 Apr 2026. ## Exploitation reported so far SOCRadar published a report on 31 Jul 2026 about a China-nexus campaign it calls SNOWLIGHT. Its timeline lists exploitation of CVE-2026-34486 against Taiwanese Tomcat hosts as "~Apr 24-29" 2026 (a range, so 24 Apr is the earliest date it gives), and exploitation of CVE-2025-24813 on 8 May 2026. SOCRadar's recovered command history shows a public Java deserialization toolkit pointed at Tomcat 9.0.116 on port 4000, delivering a download-and-run shell command that installs the SNOWLIGHT malware. SOCRadar's target table lists 98 raw entries and 37 unique hosts for this CVE, with 2 confirmed compromises by live command execution. CISA added the CVE to KEV on 4 Aug 2026, four days after the report. KEV lists ransomware campaign use as Unknown. SOCRadar describes the Tomcat flaws as separate exploitation runs, and neither Apache nor CISA publishes the steps of the chain with CVE-2025-24813. ## Why static reachability settles it Exposure needs three things at once: one of three exact Tomcat builds, an EncryptInterceptor in the cluster channel, and network reach to the receiver port. Inspecting the version, server.xml and listening sockets answers all three on a given node, so static reachability settles it. KEV and SOCRadar show real attacks, which raises the priority for any node that passes all three checks. See [static reachability](/methods/static-reachability/) for what the method proves and misses. ## Related pages - [CVE-2026-123158: OpenBSD OpenSSH, OpenSSH musl no-PAM password bypass](/ledger/cve-2026-123158/) - [CVE-2026-46300: Linux Kernel](/ledger/cve-2026-46300/) - [CVE-2026-19868: SonicWall SonicOS SSLVPN, SonicOS SSLVPN SAML bypass](/ledger/cve-2026-19868/) ## Sources - CVE-2026-34486: NVD record: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-34486 (read 19 Nov 2026) - CVE-2026-34486: CISA KEV entry (feed version 2026.11.18): https://www.cisa.gov/known-exploited-vulnerabilities-catalog?field_cve=CVE-2026-34486 (read 19 Nov 2026) - CVE-2026-34486: Apache Tomcat security announcement (vendor advisory, 9 Apr 2026): https://lists.apache.org/thread/9510k5p5zdvt9pkkgtyp85mvwxo2qrly (read 19 Nov 2026) - CVE-2026-34486: Apache Tomcat 11 security page (vendor advisory, per-branch details, report dates): https://tomcat.apache.org/security-11.html (read 19 Nov 2026) - CVE-2026-34486: Apache Tomcat 10 security page (vendor advisory): https://tomcat.apache.org/security-10.html (read 19 Nov 2026) - CVE-2026-34486: Apache Tomcat 9 security page (vendor advisory): https://tomcat.apache.org/security-9.html (read 19 Nov 2026) - CVE-2026-34486: Tomcat 11 changelog (release dates): https://tomcat.apache.org/tomcat-11.0-doc/changelog.html (read 19 Nov 2026) - CVE-2026-34486: Tomcat 10.1 changelog (release dates): https://tomcat.apache.org/tomcat-10.1-doc/changelog.html (read 19 Nov 2026) - CVE-2026-34486: Tomcat 9 changelog (release dates): https://tomcat.apache.org/tomcat-9.0-doc/changelog.html (read 19 Nov 2026) - CVE-2026-34486: Tomcat 11 cluster interceptor reference: https://tomcat.apache.org/tomcat-11.0-doc/config/cluster-interceptor.html (read 19 Nov 2026) - CVE-2026-34486: Tomcat 11 clustering guide: https://tomcat.apache.org/tomcat-11.0-doc/cluster-howto.html (read 19 Nov 2026) - CVE-2026-34486: SOCRadar research on the SNOWLIGHT campaign (exploitation and indicators): https://socradar.io/blog/snowlight-government-chinese-campaign/ (read 19 Nov 2026) - CVE-2026-34486: Red Hat CVE page: https://access.redhat.com/security/cve/CVE-2026-34486 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/ledger/cve-2026-34486/ --- # EPSS exploit prediction: what it proves and what it misses Method, read 19 Nov 2026 **The short answer.** EPSS is FIRST's daily estimate, between 0 and 1, of the probability that a CVE is exploited in the wild within the next 30 days. On 18 Nov 2026 the 37 ledger CVEs ranged from 0.0018 to 0.99999, and 16 of the 33 that CISA lists as exploited had an EPSS score below 0.10. FIRST's EPSS FAQ defines it this way: "EPSS (Exploit Prediction Scoring System) is a data-driven model that estimates the probability a vulnerability will be exploited in the wild within the next 30 days." It gives every public CVE an EPSS score between 0 and 1, updated daily, generated by Empirical Security for the EPSS Special Interest Group at FIRST. The values are calibrated: an EPSS of 0.05 "means a 5% estimated probability," so among many CVEs near 0.05 about 1 in 20 should see exploitation activity within 30 days. A second figure, the percentile, says what share of scored CVEs sit below the CVE. To run it, fetch the EPSS score for each CVE on the day you [triage](/guides/vulnerability-triage-with-evidence/), for example https://api.first.org/data/v1/epss?cve=CVE-2026-82329, which accepts a comma-separated list. The reply carries epss, percentile and date. The step that takes judgment is the cut-off. The FAQ says "there is no universal threshold" and that "a threshold of 0.10 (10%) is commonly cited in the field but carries no special authority from EPSS." It points teams to its coverage, effort and efficiency framework: pick the line your remediation capacity can sustain. The table below gives the EPSS score and percentile of all 37 ledger CVEs, fetched on 19 Nov 2026 with FIRST's EPSS date of 18 Nov 2026, beside each CVE's status in the CISA catalog. 17 have an EPSS of 0.10 or higher and 20 are lower. The 4 that are not in the catalog (CVE-2026-75604, CVE-2026-94545, CVE-2026-16723 and CVE-2026-80521) are all in the lower group, at 0.02105, 0.0108, 0.0066 and 0.0018. A 0.10 line puts the Artifactory flaw [CVE-2026-42018](/ledger/cve-2026-42018/) (0.09983) just below it and MLflow CVE-2026-64849 (0.10017) just above it, though both are in the CISA catalog. The EPSS score gives the wrong answer in four ways. It can be low for a flaw that is already exploited: CVE-2026-41709 has an EPSS of 0.0069 and has been on the CISA list since 28 Oct 2026, so a queue cut at 0.10 drops it, which the [guide to reading a vendor advisory](/guides/reading-a-vendor-advisory/) pairs with the advisory's own conditions. It moves daily, so a figure read last week may not match today's. A CVE published in the last few days has little history behind its EPSS score. And a high EPSS score belongs to the CVE, not the product you run: CVE-2026-76460 at 0.7560 is irrelevant to a team with no Cisco ISE. [Exploitable versus vulnerable](/guides/exploitable-vs-vulnerable/) shows how to settle that for your own hosts. ## EPSS for the 37 ledger CVEs | CVE | EPSS | Percentile | In CISA catalog | | --- | --- | --- | --- | | CVE-2026-109619 | 0.99999 | 0.99999 | Yes | | CVE-2026-109618 | 0.99999 | 0.99999 | Yes | | CVE-2026-21589 | 0.99664 | 0.9995 | Yes | | CVE-2026-75650 | 0.99376 | 0.9994 | Yes | | CVE-2026-34486 | 0.98503 | 0.99921 | Yes | | CVE-2026-111497 | 0.82641 | 0.99674 | Yes | | CVE-2026-76460 | 0.75599 | 0.99527 | Yes | | CVE-2026-97229 | 0.69289 | 0.9937 | Yes | | CVE-2026-102764 | 0.50694 | 0.98943 | Yes | | CVE-2026-19490 | 0.33262 | 0.98402 | Yes | | CVE-2026-9198 | 0.28658 | 0.98175 | Yes | | CVE-2026-94127 | 0.2655 | 0.98052 | Yes | | CVE-2026-123158 | 0.189 | 0.97351 | Yes | | CVE-2026-46300 | 0.1613 | 0.96989 | Yes | | CVE-2026-82329 | 0.14372 | 0.96684 | Yes | | CVE-2026-88779 | 0.11388 | 0.96077 | Yes | | CVE-2026-64849 | 0.10017 | 0.95707 | Yes | | CVE-2026-42018 | 0.09983 | 0.95695 | Yes | | CVE-2026-65660 | 0.09852 | 0.95656 | Yes | | CVE-2026-42016 | 0.08643 | 0.95213 | Yes | | CVE-2026-68820 | 0.08567 | 0.95175 | Yes | | CVE-2026-81963 | 0.07115 | 0.94366 | Yes | | CVE-2026-104286 | 0.04573 | 0.91796 | Yes | | CVE-2026-87491 | 0.03507 | 0.8942 | Yes | | CVE-2026-0322 | 0.02147 | 0.82468 | Yes | | CVE-2026-75604 | 0.02105 | 0.82112 | No | | CVE-2026-119977 | 0.01586 | 0.76089 | Yes | | CVE-2026-84869 | 0.01173 | 0.68102 | Yes | | CVE-2026-94545 | 0.01082 | 0.65868 | No | | CVE-2026-53266 | 0.00934 | 0.61458 | Yes | | CVE-2026-59822 | 0.00836 | 0.58423 | Yes | | CVE-2026-76418 | 0.00821 | 0.58002 | Yes | | CVE-2026-41709 | 0.00689 | 0.53464 | Yes | | CVE-2026-16723 | 0.00663 | 0.52431 | No | | CVE-2026-19868 | 0.00532 | 0.45663 | Yes | | CVE-2026-85880 | 0.00387 | 0.33457 | Yes | | CVE-2026-80521 | 0.00178 | 0.07411 | No | Highest EPSS first. EPSS date 18 Nov 2026; catalog release 2026.11.18. The percentile is the share of scored CVEs that sit below the CVE. ## A worked threshold decision A team that can patch 10 of these CVEs a week sets the line from its capacity, not from 0.10. In the table the tenth CVE is CVE-2026-19490 at 0.33262 and the eleventh is CVE-2026-9198 at 0.28658, so a line of 0.30 selects exactly 10, and all 10 are in the CISA catalog. The 23 catalog entries below the line, including CVE-2026-41709 at 0.00689, are not dropped: the catalog listing sends them to a second queue that runs on CISA's due dates. ## Who records the EPSS value - Vulnerability analyst: on the day of triage, writes the EPSS score, the percentile and the EPSS date into the ticket, because the value changes daily. Record: those three values. - Vulnerability manager: sets the threshold from remediation capacity and signs it, then reviews it each quarter or when capacity changes. Record: the threshold, the capacity figure it came from, and the date. - Analyst: when a CVE moves across the line on a later day, adds the new value beside the old one rather than overwriting it. ## Related pages - [Known-exploited lists: what they prove and what they miss](/methods/known-exploited/) - [Static reachability and configuration reads: what they prove and what they miss](/methods/static-reachability/) - [Runtime reachability and live-state reads: what they prove and what they miss](/methods/runtime-reachability/) ## Sources - FIRST EPSS FAQ: what EPSS is, is not, and threshold guidance: https://www.first.org/epss/faq (read 19 Nov 2026) - FIRST EPSS API, queried for the 37 ledger CVEs, EPSS date 18 Nov 2026: https://api.first.org/data/v1/epss?cve=CVE-2026-0322,CVE-2026-102764,CVE-2026-104286,CVE-2026-109618,CVE-2026-109619,CVE-2026-111497,CVE-2026-119977,CVE-2026-123158,CVE-2026-16723,CVE-2026-19490,CVE-2026-19868,CVE-2026-21589,CVE-2026-34486,CVE-2026-41709,CVE-2026-42016,CVE-2026-42018,CVE-2026-46300,CVE-2026-53266,CVE-2026-59822,CVE-2026-64849,CVE-2026-65660,CVE-2026-68820,CVE-2026-75604,CVE-2026-75650,CVE-2026-76418,CVE-2026-76460,CVE-2026-80521,CVE-2026-81963,CVE-2026-82329,CVE-2026-84869,CVE-2026-85880,CVE-2026-87491,CVE-2026-88779,CVE-2026-9198,CVE-2026-94127,CVE-2026-94545,CVE-2026-97229 (read 19 Nov 2026) - FIRST EPSS home: https://www.first.org/epss/ (read 10 Nov 2026) - FIRST EPSS article on probability and percentile: https://www.first.org/epss/articles/prob_percentile_bins (read 10 Nov 2026) - CISA KEV JSON feed, version 2026.11.18, used for the KEV counts: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/methods/exploit-prediction/ --- # Known-exploited lists: what they prove and what they miss Method, read 19 Nov 2026 **The short answer.** The CISA Known Exploited Vulnerabilities catalog lists CVEs with reliable evidence of active exploitation and a clear fix. Release 2026.11.18 holds 1,786 entries, and 33 of the 37 ledger CVEs are in it. A listing shows the flaw was exploited somewhere, not on your systems, and its dateAdded is the latest date exploitation can have started. CISA's Binding Operational Directive 22-01 says: "CISA will determine vulnerabilities warranting inclusion in the catalog based on reliable evidence that the exploit is being actively used to exploit public or private organizations by a threat actor." The catalog page lists three thresholds: an assigned CVE ID, "reliable evidence that the vulnerability has been actively exploited in the wild," and "a clear remediation action for the vulnerability, such as a vendor-provided update." The directive binds federal executive civilian agencies, who must remediate within a timeframe CISA sets. CISA's FAQ adds two cautions: "a vulnerability's exploitability is not considered as criteria for inclusion in the KEV catalog," and "we do not recommend exclusive use of the KEV Catalog as the only criterion on which organizations triage vulnerabilities." The catalog page says BOD 26-04, issued on 10 Jun 2026, carries these criteria forward. In practice a team downloads the JSON or CSV feed on a schedule and joins its cveID field to the CVE IDs in its scanner and asset inventory. Each entry carries dateAdded, dueDate, knownRansomwareCampaignUse and forensicTriage. BOD 26-04 defines the forensic-triage marker: remediation or mitigation within three days, and "carry out a forensic triage of the asset to assess whether the system is compromised." The release read on 19 Nov 2026 is version 2026.11.18, dated 18 Nov 2026 and holding 1,786 entries. New entries trigger a check for affected assets, and the dueDate gives the team a reference date for its own deadline. Two ledger entries show the range. CVE-2026-102764 is an origin-validation flaw in BeyondTrust Remote Support, rated 9.3 (CVSS 4.0 base score). BeyondTrust and Mandiant place the first exploitation at 29 Aug 2026, 47 days before disclosure on 15 Oct 2026. CISA listed it on 16 Oct 2026 with a due date of 19 Oct 2026. A team that waited for the catalog learned of it 48 days after the first attack. CVE-2026-41709, the ESX flaw rated 2.7 (CVSS base score), was listed on 28 Oct 2026 with a due date of 11 Nov 2026, 14 days out. Across the 33 ledger CVEs in the feed, 22 carry a 3-day window and 11 a 14-day window, and forensicTriage reads Yes on 19 of the 22 three-day entries and on none of the 14-day entries. The list gives the wrong answer in these cases. Absence is not safety: four ledger CVEs are absent from the 18 Nov feed, listed in the table below. Listing can lag exploitation by months: SOCRadar places exploitation of Apache Tomcat CVE-2026-34486 at about 24 to 29 Apr 2026, and CISA listed it on 4 Aug 2026, about 102 days after the start of that range. A listing says a flaw is exploited somewhere, so a listed flaw in a product you run only on an isolated segment still needs a reachability check. The KEV added and due dates for all 37 ledger CVEs are in the [ledger dataset](/datasets/). ## The four ledger CVEs not in the catalog | CVE | Product | Disclosed | CVSS base | EPSS (18 Nov 2026) | Dated exploitation in the ledger | | --- | --- | --- | --- | --- | --- | | CVE-2026-16723 | fastjson 1.x | 21 Jul 2026 | 9.0 | 0.00663 | None | | CVE-2026-75604 | Next.js on Windows | 25 Aug 2026 | 9.0 | 0.02105 | None | | CVE-2026-80521 | Linux AF_UNIX garbage collector | 26 Aug 2026 | 7.8 | 0.00178 | None | | CVE-2026-94545 | Satori and Next.js next/og | 30 Sep 2026 | 5.3 | 0.01082 | None | Status in release 2026.11.18 of the CISA feed. Absence from the catalog means CISA holds no listing for the CVE; it does not say that the flaw is safe to leave. CVE-2026-94545 is dated by its advisory publication on 30 Sep 2026; the Next.js 16.3.6 release that carries the fix came on 22 Sep 2026. ## A worked join on release 2026.11.18 The inventory is constructed; the output below is copied unchanged from the feed, release 2026.11.18 downloaded on 19 Nov 2026, where dates are written YYYY-MM-DD. inventory.csv has the columns host and cve and five rows: mail-gw-01 with CVE-2026-104286, fw-edge-02 with CVE-2026-0322, esx-lab-03 with CVE-2026-41709, orders-api with CVE-2026-16723 and tomcat-04 with CVE-2026-34486. The command pulls the four fields a team needs for each match: ``` jq -r --rawfile inv inventory.csv '($inv | gsub("\r";"") | split("\n")[1:] | map(select(length>0) | split(",")[1])) as $ids | .vulnerabilities[] | select(.cveID as $c | $ids | index($c)) | [.cveID, .dateAdded, .dueDate, .knownRansomwareCampaignUse, .forensicTriage] | @csv' known_exploited_vulnerabilities.json | sort ``` It printed: ``` "CVE-2026-0322","2026-10-14","2026-10-17","Unknown","Yes" "CVE-2026-104286","2026-10-01","2026-10-04","Unknown","Yes" "CVE-2026-34486","2026-08-04","2026-08-07","Unknown","No" "CVE-2026-41709","2026-10-28","2026-11-11","Unknown","No" ``` Four of the five hosts match. CVE-2026-16723 on orders-api returns no line because it is not in the feed, so the ticket records it as not listed, which says nothing about exposure (see the table above). The windows differ: mail-gw-01 (1 Oct to 4 Oct) and fw-edge-02 (14 Oct to 17 Oct) have 3 days and forensicTriage Yes, so each needs the patch and a compromise check by the due date. tomcat-04 (4 Aug to 7 Aug) also has 3 days but forensicTriage No, so it needs the patch only. esx-lab-03 (28 Oct to 11 Nov) has 14 days and No. All four report knownRansomwareCampaignUse as Unknown, which is not a statement that no ransomware actor uses the flaw. ## Steps, owners and how they fail - Vulnerability analyst: downloads the JSON feed on a schedule, joins cveID to the CVE IDs in the scanner and the asset inventory, and lists new entries. Record: the catalogVersion read, and for each match the dateAdded and dueDate. - Asset owner: confirms whether the matched product and version are in use and where. Record: the affected asset list in the ticket. - Vulnerability manager: sets the internal deadline, which is the dueDate or an earlier date the organization chooses. Record: the deadline and who agreed it. - Failure: a stale copy. Compare the catalogVersion with the current release before each report, and do not report from a copy older than the last listing. - Failure: inventory gaps. A listed CVE that matches no asset is recorded as unmatched, not as not affected, until the inventory is shown to cover the product. - Failure: names that do not match. The feed gives vendorProject and product as text, so a product known to the inventory under another name is missed unless the join also uses the CVE ID. ## Ledger entries it settles - [CVE-2026-41709: Broadcom VMware ESX](/ledger/cve-2026-41709/): disclosed 29 Jul 2026 - [CVE-2026-75650: Adobe Commerce and Magento Open Source](/ledger/cve-2026-75650/): disclosed 7 Sep 2026 ## Related pages - [Runtime reachability and live-state reads: what they prove and what they miss](/methods/runtime-reachability/) - [EPSS exploit prediction: what it proves and what it misses](/methods/exploit-prediction/) - [Severity scores: what they prove and what they miss](/methods/severity-score/) ## Sources - CISA BOD 22-01: https://www.cisa.gov/news-events/directives/bod-22-01-reducing-significant-risk-known-exploited-vulnerabilities (read 19 Nov 2026) - CISA KEV catalog page: three criteria and FAQ: https://www.cisa.gov/known-exploited-vulnerabilities (read 19 Nov 2026) - CISA KEV JSON feed, version 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - BeyondTrust blog on BT26-05 (CVE-2026-102764): https://www.beyondtrust.com/blog/entry/bt26-05 (read 19 Nov 2026) - Broadcom advisory for CVE-2026-41709: https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017 (read 10 Nov 2026) - CISA kev-data README: update schedule: https://raw.githubusercontent.com/cisagov/kev-data/develop/README.md (read 10 Nov 2026) - SOCRadar report on the SNOWLIGHT campaign (CVE-2026-34486): https://socradar.io/blog/snowlight-government-chinese-campaign/ (read 19 Nov 2026) - CISA BOD 26-04: Prioritizing Security Updates Based on Risk: https://www.cisa.gov/news-events/directives/bod-26-04-prioritizing-security-updates-based-risk (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/methods/known-exploited/ --- # Runtime reachability and live-state reads: what they prove and what they miss Method, read 19 Nov 2026 **The short answer.** Runtime reachability reads the state of running systems: a sensor that records which packages and functions are loaded, and direct reads of a live service's modules, settings and listening addresses. 7 of the 37 entries on 19 Nov 2026 turn on such facts, such as whether the ebtables SNAT module is in use on a Linux host (CVE-2026-53266). No standards body defines the term, so vendor documentation is the primary source. Endor Labs describes instrumented container reachability as embedding "a runtime sensor in your container image to record how your application uses the image in a real environment." Sysdig lists as a feature a focus on "the vulnerabilities tied to the packages loaded at runtime," and states that pipeline scans carry no runtime context, so a CI scan is a static snapshot while an agent on the node sees what runs. In practice a sensor is installed in the image or on the node and the workload is exercised. Endor Labs suggests profiling in "a realistic environment, such as staging, that mirrors production traffic," or when you "already have integration or end-to-end tests and want reachability to reflect those tests." It runs through Kubernetes or Docker with the endorctl container scan command and the --os-reachability flag. The output sorts each finding by whether its package was seen in use. Findings in packages never loaded move down the queue; they are not deleted. CVE-2026-53266 is an out-of-bounds write in the Linux ebtables SNAT target. The fix was committed to mainline on 1 Jun 2026, Red Hat's public date, and the kernel CNA published the CVE record on 25 Jun 2026. Backports shipped in kernels 5.10.259, 5.15.210, 6.1.176, 6.6.143, 6.12.94, 6.18.36 and 7.0.13. A version scan flags every host in the affected ranges, for example a host on 6.1.150, which lies in the 5.16 to before 6.1.176 range. The ledger record says the deciding facts are runtime ones: bridge netfilter and ebtables must be in use, with a rule using the snat target with ARP rewrite, and the attacker needs local low-privileged access. Whether a local user can install such a rule depends on capabilities such as CAP_NET_ADMIN, which the advisory does not state. CISA listed the flaw on 18 Sep 2026 with a due date of 21 Sep 2026; the ledger takes that listing as the first dated report of exploitation, so exploitation may have begun earlier. Its EPSS score on 18 Nov 2026 was 0.00934. The reads that settle it are ebtables -t nat -L as root, looking for -j snat rules with --snat-arp, and lsmod | grep -E 'ebtable_nat|ebt_snat'. On a host where the module is not loaded and no such rule exists, the version finding over-reports. The method gives a wrong 'not in use' when the sensor watched the wrong period. A library loaded by a quarterly reconciliation job is absent from a one-week profile, and the finding is demoted until the job runs. Endor Labs ties its result to the profiled usage, so a staging run that mirrors only part of production traffic yields a partial answer. The reverse error is a library that is loaded but only reachable by a caller who never meets an attacker, which loaded-at-runtime evidence cannot tell apart from one that can be attacked. The other six of the seven entries turn on the state of a running service or host rather than on loaded packages, and a package sensor does not report these. They are answered by reading the live system. The table gives the read that settles each. ## The live-state reads | Entry | The state that decides | The read that settles it | | --- | --- | --- | | Langflow, CVE-2026-9198 | The effective LANGFLOW_AUTO_LOGIN value and the listening address | env \| grep LANGFLOW_AUTO_LOGIN, then ss -ltnp \| grep -i langflow. An unset value on 1.10.0 or earlier counts as enabled; 1.10.1 is the first fixed release. | | LiteLLM, CVE-2026-59822 | The MCP server authentication type and whether the /mcp/ route is reachable | grep -n -A8 mcp_servers config.yaml for the auth_type of each server, then curl -s -o /dev/null -w '%{http_code}\n' https://PROXY_HOST/mcp/ from an untrusted network. A 401 or 403 means the route is reachable and rejects anonymous callers. | | MLflow, CVE-2026-64849 | Whether the webhooks API is open without authentication and what the server can reach outbound | ps -ef \| grep 'mlflow server' and look for --app-name basic-auth; without it the API is open. Then ss -ltnp \| grep -i mlflow for the bind address. | | Linux kernel, CVE-2026-46300 | Whether the ESP modules can load and whether unprivileged namespaces are allowed | lsmod \| grep -E 'esp4\|esp6' and sysctl user.max_user_namespaces. Red Hat documents the module block and the namespace setting as separate mitigations, so a host is exposed only while both are open. | | SharePoint, CVE-2026-65660 | The farm build and whether any site allows anonymous viewing | (Get-SPFarm).BuildVersion in the SharePoint Management Shell, compared with 16.0.5565.1001 (2016), 16.0.10417.20198 (2019) and 16.0.19725.20522 (Subscription Edition). Anonymous access: Central Administration > Application Management > Manage web applications > Authentication Providers. | | WordPress, CVE-2026-119977 | Whether a core upgrade runs and whether the staging path answers over HTTP | curl -sI https://your-site.example/wp-content/upgrade/ from outside the network: 200 or a directory listing means reachable, 403 or 404 means the condition is removed. During an upgrade, ls -la wp-content/upgrade/ shows a package-named folder with no random suffix when the unfixed code ran. | The SharePoint build read counts as a runtime read because Get-SPFarm asks the running farm which build it reports, and the anonymous-access setting is read from its live web applications. A scan of installer files or a package list can show neither. ## A worked sensor result, with constructed numbers The figures are constructed to show the shape of a result; they are not output from a product. Two bridge hosts run kernel 6.1.150, inside the affected range for CVE-2026-53266, so a version scan flags both. A sensor profiles each for 14 days. On host A, 380 of 1,040 installed packages were loaded, the ebtables modules never loaded, and no ebtables rule exists; the finding is demoted to not in use, with the window and the module list recorded. On host B the bridge netfilter and ebt_snat modules were loaded and a nat rule with the --snat-arp option exists; the finding stays at the front of the queue. Host A's finding is demoted, not deleted, so a rule added after day 14 would show on the next profile. ## Who sets the window and what it leaves - Platform or site reliability owner, with the application owner: sets the observation window so that it covers the longest job cycle of the workload (a monthly or quarterly job needs a window that includes it). Record: window start and end, workloads covered, and the sensor version. - Security analyst: reviews each demotion before the finding's deadline and checks the window against the job calendar. Record: reviewer, date and the window used. - Application owner: reports any change that adds code paths, such as a new job or a new rule, which restarts the profile. Record: the change ticket linked to the finding. ## Ledger entries it settles - [CVE-2026-119977: WordPress Core](/ledger/cve-2026-119977/): disclosed 28 Oct 2026 - [CVE-2026-46300: Linux Kernel](/ledger/cve-2026-46300/): disclosed 13 May 2026 - [CVE-2026-53266: Linux kernel](/ledger/cve-2026-53266/): disclosed 1 Jun 2026 - [CVE-2026-59822: BerriAI LiteLLM](/ledger/cve-2026-59822/): disclosed 8 Jul 2026 - [CVE-2026-64849: MLflow Tracking Server](/ledger/cve-2026-64849/): disclosed 17 Aug 2026 - [CVE-2026-65660: Microsoft SharePoint Server](/ledger/cve-2026-65660/): disclosed 11 Aug 2026 - [CVE-2026-9198: IBM Langflow OSS](/ledger/cve-2026-9198/): disclosed 2 Jul 2026 ## Related pages - [Severity scores: what they prove and what they miss](/methods/severity-score/) - [Known-exploited lists: what they prove and what they miss](/methods/known-exploited/) - [Static reachability and configuration reads: what they prove and what they miss](/methods/static-reachability/) ## Sources - Endor Labs instrumented container reachability: https://docs.endorlabs.com/scan/containers/instrumented-reachability/index (read 19 Nov 2026) - Sysdig vulnerability management documentation: https://docs.sysdig.com/en/sysdig-secure/vulnerability-management/ (read 19 Nov 2026) - CISA KEV JSON feed, version 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - FIRST EPSS API for CVE-2026-53266, EPSS date 18 Nov 2026: https://api.first.org/data/v1/epss?cve=CVE-2026-53266 (read 19 Nov 2026) - NVD record for CVE-2026-53266: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-53266 (read 10 Nov 2026) - Upwind home page (vendor claim): https://www.upwind.io/ (read 10 Nov 2026) - Oligo home page (vendor claim): https://www.oligo.security/ (read 10 Nov 2026) - Onit platform page (vendor claim): https://onit.security/platform/ (read 10 Nov 2026) - Zafran home page (vendor claim): https://www.zafran.io/ (read 10 Nov 2026) - Astelia platform page (vendor claim): https://www.astelia.io/platform (read 10 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/methods/runtime-reachability/ --- # Severity scores: what they prove and what they miss Method, read 19 Nov 2026 **The short answer.** A CVSS base score is a number from 0 to 10 that FIRST's open framework assigns to a flaw's intrinsic characteristics and bands from Low to Critical. Of the 37 ledger CVEs read on 19 Nov 2026, 16 have a base score of 9.0 or higher, and the lowest, 2.7, is on the CISA exploited list. FIRST's CVSS v4.0 specification calls CVSS "an open framework for communicating the characteristics and severity of software vulnerabilities." Its Base group "represents the intrinsic qualities of a vulnerability that are constant over time and across user environments." The user guide is blunt about the limit: Base scores "are designed to measure the severity of a vulnerability and should not be used alone to assess risk." The specification's Table 22 maps a CVSS score to a rating: None 0.0, Low 0.1 to 3.9, Medium 4.0 to 6.9, High 7.0 to 8.9, Critical 9.0 to 10.0. In practice the number arrives with the [advisory](/guides/reading-a-vendor-advisory/). The CNA, the vendor or NVD publishes a vector string, for example CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, which gives a base score of 9.8 for [CVE-2026-82329](/ledger/cve-2026-82329/). A scanner shows the CVSS score, and the team turns the rating into a deadline. The specification says Base values "are combined with default values that assume the highest severity for Threat and Environmental metrics," and the user guide adds that "the application of Environmental and Threat metrics is the responsibility of the CVSS consumer." A CVSS score computed from the Base metrics alone is written CVSS-B. The ledger mixes v3.1 and v4.0 vectors, which use different equations, so two base scores from different versions are not on one scale. CVE-2026-41709 shows the gap. Broadcom rates the VMware ESX insufficient-logging flaw 2.7 (CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:U/C:N/I:L/A:N), a Low rating that reflects an attacker who already holds ESX administrator privileges and can skip some log entries. At 2.7 it has the lowest base score of the 37 ledger CVEs. CISA added it to the KEV catalog on 28 Oct 2026 with a due date of 11 Nov 2026. In the other direction, CVE-2026-94545 carries two CVSS 4.0 scores for one flaw: 9.5 on the Next.js advisory and 5.3 on the Satori advisory (GitHub published the Next.js advisory on 30 Sep 2026 and the Satori advisory on 27 Oct 2026). A base score gives the wrong answer in three situations: when exploitation is already happening and the flaw has a low CVSS score, as with CVE-2026-41709 above; when the vendor's own context moves the number only slightly, as with CVE-2026-0322, where Palo Alto's base score falls from 8.6 to 8.5 when the management interface is restricted to a jump box; and when two advisories assess the same flaw in different products, as with CVE-2026-94545. Adding Threat metrics, which the specification says rest on threat intelligence the consumer supplies, narrows the first gap but leaves the other two. ## The 37 ledger CVEs by rating | Rating | Base score | CVEs | Of which in the CISA catalog | | --- | --- | --- | --- | | Critical | 9.0 to 10.0 | 16 | 14 | | High | 7.0 to 8.9 | 19 | 18 | | Medium | 4.0 to 6.9 | 1 | 0 | | Low | 0.1 to 3.9 | 1 | 1 | CVSS base scores as recorded in the ledger on 19 Nov 2026; CVSS 3.1 and CVSS 4.0 scores are counted together. ## A worked rescoring of CVE-2026-0322 The base vector from Palo Alto is CVSS:4.0/AV:N/AC:L/AT:N/PR:H/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N, a CVSS base score of 8.6. Palo Alto reports the exploit maturity as attacked, so the Threat metric is E:A. Because an unset Threat metric already assumes E:A, the vector with E:A is still 8.6: the exploitation evidence confirms the base score and does not raise it. Now the Environmental step. The management interface is reachable only from a jump box on the adjacent network, so set Modified Attack Vector to adjacent. The vector becomes ...SA:N/E:A/MAV:A. A CVSS score that combines Base, Threat and Environmental metrics is written CVSS-BTE, and this one is 8.5, which matches Palo Alto's own figure for the jump-box case. If the firewall is a lab device and its owner also sets the security requirements CR, IR and AR to low, the CVSS score falls to 7.0. Had the Threat value been E:P (proof-of-concept only), the CVSS score with the jump box would be 7.1, and with E:U (unreported) 5.9. All four figures were computed with the CVSS 4.0 equations; the Threat value is a judgment about exploitation, which is why the KEV listing matters to it. ## Who applies the metrics and where they are recorded - Asset owner: proposes the Environmental values (Modified base metrics and the CR, IR and AR requirements) from what the device does and who can reach it. - Security analyst: sets the Threat value from named evidence, such as a CISA listing or the vendor's exploit maturity, and computes the CVSS score. - Record: the full vector string, the resulting CVSS score, the date, and the evidence behind each Threat and Environmental value, stored in the ticket beside the vendor's base vector. ## Related pages - [Static reachability and configuration reads: what they prove and what they miss](/methods/static-reachability/) - [Runtime reachability and live-state reads: what they prove and what they miss](/methods/runtime-reachability/) - [EPSS exploit prediction: what it proves and what it misses](/methods/exploit-prediction/) ## Sources - FIRST CVSS v4.0 specification, version 1.2: https://www.first.org/cvss/v4.0/specification-document (read 19 Nov 2026) - FIRST CVSS v4.0 user guide: Base score measures severity, not risk: https://www.first.org/cvss/v4.0/user-guide (read 19 Nov 2026) - CISA KEV JSON feed, version 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - Broadcom advisory for CVE-2026-41709: https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/38017 (read 19 Nov 2026) - Next.js advisory for CVE-2026-94545 (CVSS 4.0 9.5): https://github.com/vercel/next.js/security/advisories/GHSA-vcvr-r3jv-pc5j (read 19 Nov 2026) - Satori advisory for CVE-2026-94545 (CVSS 4.0 5.3): https://github.com/vercel/satori/security/advisories/GHSA-wx4j-mvgx-mqwp (read 19 Nov 2026) - Palo Alto Networks advisory for CVE-2026-0322: https://security.paloaltonetworks.com/CVE-2026-0322 (read 19 Nov 2026) - NVD record for CVE-2026-41709: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-41709 (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/methods/severity-score/ --- # Static reachability and configuration reads: what they prove and what they miss Method, read 19 Nov 2026 **The short answer.** Static reachability reads facts that need no attack: a call graph from source code or build output that shows whether the team's code reaches a vulnerable function, and, for appliances and hosts, the version and configuration settings. 14 of the 37 entries on 19 Nov 2026 turn on such reads. A no-path label rules out only the code analyzed. No standards body defines the term, so the vendors' documentation does. Snyk describes reachability analysis as "identifying whether your application is calling a code element (for example, functions, classes, modules, and more) related to the vulnerability." Endor Labs labels a finding a Reachable Function when "there is a path from the developer-written code to a vulnerable function." Semgrep takes a pattern approach: "If there's a code pattern in the codebase that matches the vulnerability definition, the finding is flagged as reachable," and a finding is conditionally reachable when it "can be exploited when specific conditions are met." The tool first reads the manifest or lockfile for package versions. It then builds the call graph of the application and of its dependencies and checks for a path to the code element that the vulnerability record names (Snyk uses its DeepCode engine for this; Semgrep matches rule patterns). It runs in CI on each merge or on a schedule, and the result is a label on each finding, with the path shown for reachable ones. Support is per language and package manager, and a vulnerability only gets a result if someone has mapped the CVE to a function or pattern. Endor Labs also offers a coarser dependency-level label, which says only that the package is imported somewhere. CVE-2026-16723, the fastjson 1.x flaw disclosed on 21 Jul 2026, has four preconditions in the vendor advisory, and all four can be read from a repository and its build: fastjson 1.2.68 to 1.2.83 in the lockfile (the fix is 1.2.84); packaging as a Spring Boot executable fat-jar, since the advisory says plain jar, uber-jar and Tomcat or Jetty WAR deployments do not meet the trigger condition; AutoType and SafeMode at their defaults, both off; and a call path from attacker-supplied JSON to JSON.parse or JSON.parseObject. An application on 1.2.79 built as a Tomcat WAR sits in the version range yet fails the packaging test. The CVE is rated 9.0 (CVSS 3.1 base score), is not in the CISA catalog as of 18 Nov 2026, and has an EPSS score of 0.00663. CVE-2026-94545 is a data-flow variant: search for next/og ImageResponse imports on the Node.js runtime, then check whether request-derived values flow into the SVG tree. Snyk spells out the first failure mode: static analysis "can show that a vulnerability or code element can be reached through at least one execution path. However, just because there is no evidence of this does not mean that the element cannot be reached." Reflection, dynamic loading and unusual control flow produce false 'no path' results, and Snyk notes that a finding marked as having no path found can change to Reachable as the engine improves. The opposite error also occurs. CVE-2026-75604 affects Next.js only on Windows-hosted servers, so a call path into the incremental cache code is real for every in-range application yet harmless on a Linux host, and source code does not say which host a build will run on. Of the 14 entries filed here, five are questions about code or a build: fastjson (CVE-2026-16723), the two Next.js entries (CVE-2026-75604, CVE-2026-94545), the Node.js Brotli decoder (CVE-2026-111497) and Tomcat (CVE-2026-34486, which also needs server.xml and the listening sockets). The other nine are version and setting reads from an appliance or host: NetScaler (CVE-2026-19490, CVE-2026-88779), FortiMail, SonicOS, ASA and FTD, BIG-IP, BeyondTrust Remote Support, OpenSSH and Atlassian. Call-graph tools do not scan appliance firmware, so these are scripted reads of the version and settings that the advisory lists, such as show ns version and a grep of the running configuration for NetScaler. Both kinds check a fact without running an attack. The [Methods Matrix](/matrix/) sets it beside the other six methods. ## A worked tool result for CVE-2026-16723, with constructed output The output below is constructed to show the shape of a result; it is not captured from a product. A service pins fastjson 1.2.79 in its lockfile. The tool reports: finding CVE-2026-16723, label Reachable Function, path OrderController.create to JsonUtil.read to com.alibaba.fastjson.JSON.parseObject. The label confirms two of the four preconditions, the version (1.2.68 to 1.2.83) and the call path. The tool does not report the packaging, so the engineer reads the build file: if it produces a Tomcat WAR, the advisory says the trigger condition is not met, and the finding is recorded as reachable but not exploitable in this build. If it produces a Spring Boot fat-jar, the engineer checks that AutoType and SafeMode are at their defaults and upgrades to 1.2.84. ## A worked appliance read for CVE-2026-88779, with constructed output The output is constructed to show the shape of a read; it is not captured from an appliance. Citrix's bulletin CTX697174 lists NetScaler ADC and Gateway 14.1 as affected before 14.1-73.41 and applies the flaw to a SAML service provider or identity provider configuration. The [ledger entry](/ledger/cve-2026-88779/) gives the commands. Take an appliance that reports build 73.38 and has one samlAction line in its running configuration and no samlIdPProfile line. Build 73.38 is before 14.1-73.41, so the version condition holds. The samlAction line makes it a SAML service provider, which meets the configuration condition, and the empty samlIdPProfile result adds nothing, because either hit is enough. The third condition, that an attacker can reach the SAML endpoint, is not on the device: check the published addresses of the virtual servers that use the action against the firewall rules. The record is the three outputs, the date and the node name, repeated on every node of an HA pair. The fix is 14.1-73.41 or later. ## Steps, owners and the record each leaves - Application security engineer: runs the tool in CI on each merge or on a schedule. Record: the label, the call path, the commit hash and the lockfile version for each finding. Failure: a language or package manager the tool does not support returns no finding, which reads as clean. The engineer lists unsupported manifests and records them as not analyzed. - Application owner: reads the preconditions in the advisory that the tool does not check (packaging, defaults, host operating system) from the build and configuration. Record: one line per precondition, holding or failing. Failure: reading the repository instead of the deployed build, so a precondition set only at deploy time is missed. The owner reads the artifact that ships. - Application security engineer: reviews every no path found result for dynamic loading or reflection before closing it. Record: reviewer, date and the code searched. Failure: closing on a label the engine later revises. The engineer re-runs closed findings after each engine update. - Appliance or network owner: reads the version and the settings the advisory names from the device and saves the output. Record: the command, the output and the date. Failure: reading a saved configuration file that differs from the running one, or reading before a later change. The owner reads the running configuration and repeats the read after any change window. ## Ledger entries it settles - [CVE-2026-102764: BeyondTrust Remote Support](/ledger/cve-2026-102764/): disclosed 15 Oct 2026 - [CVE-2026-104286: Fortinet FortiMail](/ledger/cve-2026-104286/): disclosed 1 Oct 2026 - [CVE-2026-111497: Node.js Brotli decoder](/ledger/cve-2026-111497/): disclosed 21 Oct 2026 - [CVE-2026-123158: OpenBSD OpenSSH](/ledger/cve-2026-123158/): disclosed 5 Nov 2026 - [CVE-2026-16723: Alibaba fastjson 1.x](/ledger/cve-2026-16723/): disclosed 21 Jul 2026 - [CVE-2026-19490: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-19490/): disclosed 19 Aug 2026 - [CVE-2026-19868: SonicWall SonicOS SSLVPN](/ledger/cve-2026-19868/): disclosed 5 Nov 2026 - [CVE-2026-21589: Atlassian Data Center products](/ledger/cve-2026-21589/): disclosed 5 Oct 2026 - [CVE-2026-34486: Apache Software Foundation Apache Tomcat](/ledger/cve-2026-34486/): disclosed 9 Apr 2026 - [CVE-2026-75604: Vercel Next.js](/ledger/cve-2026-75604/): disclosed 25 Aug 2026 - [CVE-2026-76418: Cisco Secure Firewall ASA and FTD](/ledger/cve-2026-76418/): disclosed 28 Oct 2026 - [CVE-2026-88779: Citrix NetScaler ADC and Gateway](/ledger/cve-2026-88779/): disclosed 3 Oct 2026 - [CVE-2026-94127: F5 BIG-IP APM](/ledger/cve-2026-94127/): disclosed 22 Sep 2026 - [CVE-2026-94545: Vercel Satori and Next.js](/ledger/cve-2026-94545/): disclosed 30 Sep 2026 ## Related pages - [EPSS exploit prediction: what it proves and what it misses](/methods/exploit-prediction/) - [Severity scores: what they prove and what they miss](/methods/severity-score/) - [Known-exploited lists: what they prove and what they miss](/methods/known-exploited/) ## Sources - Snyk reachability analysis documentation, including limits: https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis (read 19 Nov 2026) - Endor Labs reachability analysis overview: https://docs.endorlabs.com/scan/sca/reachability-analysis/index (read 19 Nov 2026) - Semgrep Supply Chain overview: https://semgrep.dev/docs/semgrep-supply-chain/overview (read 19 Nov 2026) - Endor Labs call graphs: https://docs.endorlabs.com/scan/sca/call-graphs/index (read 10 Nov 2026) - Backline product page (vendor claim): https://backline.ai/product/ (read 10 Nov 2026) - fastjson2 wiki advisory for CVE-2026-16723: https://github.com/alibaba/fastjson2/wiki/Security-Advisory:-Remote-Code-Execution-in-fastjson-1.2.68%E2%80%931.2.83 (read 10 Nov 2026) - Next.js advisory for CVE-2026-94545: https://github.com/vercel/next.js/security/advisories/GHSA-vcvr-r3jv-pc5j (read 10 Nov 2026) - NVD record for CVE-2026-75604: https://services.nvd.nist.gov/rest/json/cves/2.0?cveId=CVE-2026-75604 (read 10 Nov 2026) - Citrix security bulletin CTX697174 (NetScaler ADC and Gateway): https://support.citrix.com/external/article/CTX697174/netscaler-adc-and-netscaler-gateway-secu.html (read 19 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/methods/static-reachability/ --- # The Methods Matrix: what each method needs, proves and misses Methods Matrix, read 19 Nov 2026 **The short answer.** Seven methods compared on the question each answers, what it needs, proves and misses, who offers it and how many ledger entries it settles. Static reachability settles 14 ledger CVEs, attack-path analysis 11, runtime reachability 7, agent investigation 3 and known-exploited lists 2. The same rows are on the [datasets page](/datasets/) as CSV and JSON. | Method | Question it answers | Needs | Proves | Misses | Offered by | Ledger entries it settles (of 37) | | --- | --- | --- | --- | --- | --- | --- | | Agent investigation | What do the configuration, code, controls and logs say about this one finding? | Read access to the cloud, code, logs and security controls involved, the scanner findings to investigate, and context such as owners and business criticality. | One result per finding, in the label set that fits the question: reachable, not reachable or unknown; compromised, no evidence found or unknown; or the hosts at or above and below a fixed build. Each result comes with the configuration, code, controls, logs or paths it read, including what blocks a path and why a finding was closed. | Anything outside the access granted, and evidence that no longer exists, such as rotated logs. A wrong 'not reachable' conclusion is caught only by the human review of closures. | Astelia, Averlon, Backline, Cogent, Dux, Maze, Onit, Qualys, Rival Security, Seemplicity, Vicarius | 3 | | Attack-path analysis | Can an attacker reach the affected asset from outside? | Read access to cloud, identity and network configuration (usually through agentless connectors), vulnerability and misconfiguration findings, and asset context such as internet exposure and data sensitivity. | That a modeled route exists, or none does, from an external entry point through exposure, identity and network controls to the affected asset, and which single fix cuts the most routes. | Whether the exploit works against your configuration, whether the vulnerable function is reachable in code, and anything the graph lacks. | Astelia, Averlon, Microsoft, Rapid7, Tenable, Wiz, XM Cyber, Zafran | 11 | | [EPSS exploit prediction](/methods/exploit-prediction/) | How likely is exploitation activity in the next 30 days? | A CVE identifier, the current EPSS score and percentile from FIRST's daily file or API, and a threshold your team chooses. | FIRST's estimate, on a scale of 0 to 1, of the probability that exploitation activity is observed for this CVE in the next 30 days. | Whether your instance is affected or reachable, which controls protect it, and how much damage a successful attack would cause. | Empirical Security, FIRST | 0 | | [Known-exploited lists](/methods/known-exploited/) | Has exploitation been seen, and does CISA set a due date? | A CVE identifier, a current copy of the catalog (CISA publishes JSON and CSV feeds) and an inventory that maps CVEs to the products you run. | That CISA holds reliable evidence of exploitation in the wild for this CVE and that a clear remediation action exists. | Flaws exploited but not yet listed, the manner and targets of the exploitation, and whether your own instance is reachable or protected. | CISA | 2 | | [Runtime reachability and live-state reads](/methods/runtime-reachability/) | Is the vulnerable code loaded or the risky setting on in the running host? | A sensor or agent on the workloads, hosts or containers (or instrumented test runs), software inventory and vulnerability data for each workload, and an observation window that exercises real behavior. For a live service, read access to its effective configuration, loaded modules and listening addresses. | That the vulnerable package, module or function is present and loaded or executed in a workload that is running, or that the running service's effective setting or listening address meets the condition in the CVE record. | Code paths that did not run while the sensor watched, who outside the host can reach a loaded package (only a read of the listening address or a request from an untrusted network settles that), compensating controls and any workload without a sensor. | Astelia, Endor Labs, Oligo, Onit, Sysdig, Upwind, Zafran | 7 | | [Severity scores](/methods/severity-score/) | How severe is the flaw in the abstract? | A CVE or advisory identifier and the published vector and base score from the CNA, vendor or NVD. Threat and Environmental values, if used, come from your team or from automation. | How severe the flaw is in the abstract: attack vector, complexity, privileges required and impact on the vulnerable system, on a scale of 0 to 10. | Whether anyone is exploiting the flaw, how exposed your instance is, what controls sit in front of it and whether the vulnerable code runs. | FIRST, NVD | 0 | | [Static reachability and configuration reads](/methods/static-reachability/) | Is the affected version, package or code path in my build? | Source code or build artifacts in a supported language, dependency manifests or lockfiles, and per-CVE data that names the vulnerable functions or code patterns. For appliances and hosts, read access to the version output and running configuration. | That a call path exists from your code to the vulnerable function, shown step by step where one is found, and that the version, packaging and configuration facts in the CVE record hold or fail in the build. | Network exposure, identity, controls, whether the code is deployed and running, and any setting changed after the read. | Backline, Endor Labs, Semgrep, Snyk | 14 | Alphabetical by method name. The last column counts the 37 ledger entries by the method each record names as settling it: 14 + 11 + 7 + 3 + 2 + 0 + 0 = 37. EPSS exploit prediction and Severity scores settle none of them, because neither says anything about one instance. ## Related pages - [Exploitability Ledger: the data, downloadable](/datasets/) - [Glossary](/glossary/) ## Sources - Onit blog on reachable vulnerabilities, 6 Nov 2026 (vendor claim): https://onit.security/blog/reachable-vulnerabilities/ (read 19 Nov 2026) - Qualys agentic AI page (vendor claim): https://www.qualys.com/apps/agentic-ai (read 19 Nov 2026) - Astelia platform page (vendor claim): https://www.astelia.io/platform (read 10 Nov 2026) - Averlon platform page (vendor claim): https://www.averlon.ai/platform (read 10 Nov 2026) - Backline product page (vendor claim): https://backline.ai/product/ (read 10 Nov 2026) - Cogent AI engine page (vendor claim): https://www.cogent.com/ai/cogent-ai-engine (read 10 Nov 2026) - Dux home page (vendor claim): https://dux.io/ (read 10 Nov 2026) - Maze cloud platform page (vendor claim): https://mazehq.com/platform/cloud (read 10 Nov 2026) - Rival Security home page (vendor claim): https://rival.security/ (read 10 Nov 2026) - Seemplicity AI Analysts announcement (vendor claim): https://seemplicity.ai/news/ai-analysts-autonomous-vulnerability-response/ (read 10 Nov 2026) - Vicarius home page (vendor claim): https://www.vicarius.io/ (read 10 Nov 2026) - Microsoft Learn: Defender for Cloud attack paths: https://learn.microsoft.com/en-us/azure/defender-for-cloud/how-to-manage-attack-path (read 19 Nov 2026) - XM Cyber glossary: attack path analysis: https://xmcyber.com/glossary/what-is-attack-path-analysis/ (read 19 Nov 2026) - Wiz Academy: attack path analysis: https://www.wiz.io/academy/attack-path-analysis (read 10 Nov 2026) - Tenable One page (vendor claim): https://www.tenable.com/products/tenable-one (read 10 Nov 2026) - Rapid7 Exposure Command page (vendor claim): https://www.rapid7.com/products/command/pricing/ (read 10 Nov 2026) - Zafran Attack Chains announcement, 3 Aug 2026 (vendor claim): https://www.zafran.io/resources/zafran-launches-attack-chain-killswitch-in-collaboration-with-google-threat-intelligence (read 10 Nov 2026) - CISA KEV JSON feed, version 2026.11.18: https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json (read 19 Nov 2026) - FIRST EPSS FAQ: what EPSS is, is not, and threshold guidance: https://www.first.org/epss/faq (read 19 Nov 2026) - FIRST EPSS home: https://www.first.org/epss/ (read 10 Nov 2026) - FIRST EPSS article on probability and percentile: https://www.first.org/epss/articles/prob_percentile_bins (read 10 Nov 2026) - CISA BOD 22-01: https://www.cisa.gov/news-events/directives/bod-22-01-reducing-significant-risk-known-exploited-vulnerabilities (read 19 Nov 2026) - CISA KEV catalog page: three criteria and FAQ: https://www.cisa.gov/known-exploited-vulnerabilities (read 19 Nov 2026) - Endor Labs instrumented container reachability: https://docs.endorlabs.com/scan/containers/instrumented-reachability/index (read 19 Nov 2026) - Sysdig vulnerability management documentation: https://docs.sysdig.com/en/sysdig-secure/vulnerability-management/ (read 19 Nov 2026) - Upwind home page (vendor claim): https://www.upwind.io/ (read 10 Nov 2026) - Oligo home page (vendor claim): https://www.oligo.security/ (read 10 Nov 2026) - Onit platform page (vendor claim): https://onit.security/platform/ (read 10 Nov 2026) - Zafran home page (vendor claim): https://www.zafran.io/ (read 10 Nov 2026) - FIRST CVSS v4.0 specification, version 1.2: https://www.first.org/cvss/v4.0/specification-document (read 19 Nov 2026) - FIRST CVSS v4.0 user guide: Base score measures severity, not risk: https://www.first.org/cvss/v4.0/user-guide (read 19 Nov 2026) - Snyk reachability analysis documentation, including limits: https://docs.snyk.io/manage-risk/prioritize-issues-for-fixing/reachability-analysis (read 19 Nov 2026) - Endor Labs reachability analysis overview: https://docs.endorlabs.com/scan/sca/reachability-analysis/index (read 19 Nov 2026) - Semgrep Supply Chain overview: https://semgrep.dev/docs/semgrep-supply-chain/overview (read 19 Nov 2026) - Endor Labs call graphs: https://docs.endorlabs.com/scan/sca/call-graphs/index (read 10 Nov 2026) Last reviewed 19 Nov 2026. Canonical: https://reachabilityanalysis.com/matrix/