CISA Adds Four Exploited Flaws to KEV: Axis, REDCap, LiteLLM and Synology DSM

by | Nov 19, 2026 | Tutorials | 0 comments

CISA added four vulnerabilities to its Known Exploited Vulnerabilities catalog on 18 November, citing evidence of active exploitation. They cover Synology DiskStation Manager, BerriAI LiteLLM, Apache Axis and Vanderbilt University’s REDCap, and the reasons behind the additions range from cryptocurrency miners to ransomware to stolen AI inference keys.

What was added

  • CVE-2026-90436 (CVSS 9.8) — unsafe reflection in Apache Axis, allowing an unauthenticated attacker to execute arbitrary code.
  • CVE-2026-90817 (CVSS 9.8) — code injection in REDCap, allowing an unauthenticated attacker to execute code on the server.
  • CVE-2026-40217 (CVSS 8.8) — unprotected alternate channel in LiteLLM, allowing remote code execution through bytecode rewriting at the /guardrails/test_custom_code URI.
  • CVE-2024-10441 (CVSS 9.8) — improper encoding or escaping of output in Synology DSM, allowing remote code execution.

Apache Axis: no patch, and it sits inside Confluence

Axis is the awkward one, because Axis 1.x is end-of-life and there is no official fix. The advice is to migrate to another SOAP engine such as Axis 2 for Java, or to build Axis with the patch from a 12 July 2026 commit applied. The KEV entry follows reports that the flaw is being chained with the 4EverAble2Shell vulnerabilities in N-able N-central to install cryptocurrency miners.

Rapid7 also documented exploitation against Atlassian Confluence Data Center, which bundles the Axis library for its XML-RPC and SOAP APIs. Atlassian’s caveat is important: the instance must have the Remote API enabled, a setting that has been off by default for new installations since Confluence 3.0 and was removed entirely in Confluence 10.0.

The timeline in that intrusion is worth reading. Attackers ran shell commands and fetched scripts from temporary Pinggy and Cloudflare Tunnel domains. Just over three minutes after the first execution they were authenticating to MySQL with Confluence’s own database credentials, then querying page bodies, application links and attachments for anything named like credentials, tokens, VPN configurations or private keys. The scripts, likely AI-generated, parsed Office documents in the attachment directory and hunted card numbers with a Luhn check. They mapped internal services, queried the Atlassian Crowd API for users and groups, and attempted AS-REP Roasting. That last step failed, but they walked away with a service account credential and queried a production cloud data warehouse. About twelve hours after the first request, someone signed in to Microsoft 365 as an integration service account over legacy authentication from 192.238.253[.]176.

REDCap: web shell, then a disguised miner

CISA’s description of CVE-2026-90817 is specific about the route: crafted HTTP requests that reach an unintended controller route from a public survey context, then supply a crafted file-path or stream parameter during data import. Exploitation needs knowledge of a valid public survey hash. Safe Security reported seeing an unknown threat actor stage a transient web shell, load an in-memory loader that pulled XMRig disguised as image files, rename the miner to look like syslog-ng, and persist through cron and an added SSH key.

LiteLLM: LLMjacking

The LiteLLM flaw was added after a Google Threat Intelligence Group report this week describing attackers using it to turn limited-access inference keys into full administrative credentials and establish persistence. The campaigns, grouped under the LLMjacking label, targeted financial services, retail, transportation and technology organisations across North America, South America and Asia Pacific.

Synology: ransomware on NAS

Synology disclosed CVE-2024-10441 earlier this month as exploited in ransomware attacks against NAS devices, leading to C3RB3R ransomware deployment. It is the oldest CVE of the four and a reminder that NAS boxes rarely get the patch attention of a public web server.

What to do

  • Federal agencies have until 21 November 2026 for the DSM, REDCap and Axis flaws, and until 2 December for LiteLLM.
  • Patch REDCap and DSM now. For Axis, migrate off 1.x or rebuild with the July commit; disabling Confluence’s Remote API where it is not needed removes the exposure Rapid7 saw.
  • Treat LiteLLM guardrail endpoints as internet-facing code execution. Restrict /guardrails/test_custom_code, rotate inference keys, and audit for admin credentials you did not create.
  • Hunt the post-exploitation patterns: outbound traffic to tunnel services, database logins from application hosts, and SSH keys or cron entries you cannot account for.

One caveat on the catalog itself: CISA still lists known ransomware campaign use as “Unknown” for all four entries, even though Synology tied its flaw to C3RB3R. The KEV flag means exploitation is confirmed, not that the full picture is public.

Sources: CISA Known Exploited Vulnerabilities catalog, The Hacker News, Rapid7, Atlassian, Safe Security and Google Threat Intelligence Group reporting as cited by The Hacker News.

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *

avatart

About

I’m a System Administrator and a developer with 10+ years of experience.