CISA flags four actively exploited flaws in Axis, REDCap, LiteLLM, and Synology
CISA added CVE-2026-90436, CVE-2026-90817, CVE-2026-40217, and CVE-2024-10441 to its KEV catalog due to active ransomware, cryptomining, and LLMjacking campaigns.
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) added four vulnerabilities to its Known Exploited Vulnerabilities catalog on November 19, 2026. The update covers critical flaws in Apache Axis, Vanderbilt University REDCap, BerriAI LiteLLM, and Synology DiskStation Manager that are currently being leveraged in active attacks.
What happened
CISA cited evidence of active exploitation for all four entries. The vulnerabilities enable remote code execution or privilege escalation across diverse infrastructure components, from legacy SOAP libraries to modern large language model gateways. Federal agencies are required to apply fixes or mitigations by November 21, 2026, for the Axis, REDCap, and Synology flaws, and by December 2, 2026, for the LiteLLM issue.
The Apache Axis vulnerability, CVE-2026-90436, has been observed in conjunction with other exploits to deploy cryptocurrency miners. Rapid7 detected threat actors using this flaw against Atlassian Confluence Data Center instances where the Remote API feature was enabled. The attackers executed shell commands, fetched scripts from temporary tunnel domains, and began enumerating the environment within minutes. They accessed MySQL databases using Confluence’s own credentials, searched for payment card numbers using Luhn checks, and attempted Kerberos-based credential harvesting.
For REDCap, CVE-2026-90817 allows unauthenticated attackers to execute arbitrary code if they possess a valid public survey hash. Safe Security reported that threat actors used this vector to stage transient web shells and load XMRig miners disguised as image files. In the AI sector, CVE-2026-40217 in LiteLLM enables attackers to rewrite bytecode via an unprotected URI, elevating limited inference keys to full administrative access. Google Threat Intelligence Group linked this activity to LLMjacking campaigns targeting financial, retail, and technology sectors globally. Finally, Synology’s CVE-2024-10441 is being used to deploy C3RB3R ransomware on network-attached storage devices.
How it works
These vulnerabilities exploit different layers of the application stack. CVE-2026-90436 is an unsafe reflection flaw in Apache Axis 1.x, a library that has reached end-of-life status. Because no official patch exists for the EoL version, attackers target systems still running legacy SOAP APIs. In Confluence environments, this allows unauthenticated remote code execution if the XML-RPC and SOAP interfaces are active.
CVE-2026-40217 in LiteLLM involves an unprotected alternate channel at the /guardrails/test_custom_code URI. Attackers use this endpoint to perform bytecode rewriting, effectively bypassing security guardrails and escalating privileges. This mechanism allows them to transform low-privilege inference keys into administrative credentials, granting persistent control over the LLM infrastructure.
Key details
- CVE-2026-90436: CVSS 9.8; affects Apache Axis 1.x; exploited for cryptomining and data theft via Confluence.
- CVE-2026-90817: CVSS 9.8; affects REDCap; requires a valid public survey hash for code injection.
- CVE-2026-40217: CVSS 8.8; affects BerriAI LiteLLM; enables LLMjacking via bytecode rewriting at a specific URI.
- CVE-2024-10441: CVSS 9.8; affects Synology DSM; leads to C3RB3R ransomware deployment on NAS devices.
- Mitigation deadlines: November 21, 2026, for Axis, REDCap, and Synology; December 2, 2026, for LiteLLM.
- Axis status: Apache Axis 1 is end-of-life; users must migrate to Axis 2/Java or apply a manual patch from July 12, 2026.
Why it matters
For engineering teams, these alerts highlight the risks of maintaining end-of-life dependencies and misconfigured AI gateways. The Axis vulnerability demonstrates how legacy libraries embedded in modern platforms like Confluence can become critical attack vectors. Since Axis 1 is no longer maintained, organizations cannot rely on vendor patches and must either migrate their SOAP engines or apply community-sourced fixes immediately.
The LiteLLM and REDCap incidents underscore the importance of securing alternative channels and public-facing endpoints. In AI infrastructure, exposing test or guardrail endpoints without proper authentication allows attackers to hijack model access and escalate privileges. For data-intensive applications like REDCap, even seemingly minor information leaks, such as a survey hash, can lead to full server compromise and cryptomining operations.
What you can do
- Audit your Confluence instances to ensure the Remote API (XML-RPC & SOAP) is disabled if not strictly required.
- Migrate away from Apache Axis 1.x to Apache Axis 2/Java or apply the July 12, 2026, patch commit if migration is not immediately possible.
- Review LiteLLM configurations to restrict access to the /guardrails/test_custom_code URI and rotate any potentially compromised inference keys.
- Verify that REDCap public surveys are configured securely and monitor for unusual file imports or web shell activity.
- Update Synology DSM to the latest version to address the improper encoding vulnerability and check for signs of C3RB3R ransomware.
- Implement network monitoring for outbound connections to temporary tunnel services like Pinggy or Cloudflare Tunnels from internal servers.



