CISA Adds Four Critical Flaws Being Exploited in Ransomware & LLMjacking Attacks

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has officially added four vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog due to confirmed misuse in attacks targeting ransomware deployment, cryptomining, and LLMjacking campaigns. These flaws affect products from Synology, LiteLLM, Apache Axis, and REDCap, and pose risks ranging from remote code execution to unauthorized control over machine learning workloads.

High-Impact Flaws Now Actively Exploited

Here’s a breakdown of the four newly added vulnerabilities, their impact scores, and how they’re being exploited:

  • Apache Axis – CVE-2026-90436 (CVSS 9.8): This flaw allows an unauthenticated attacker to execute arbitrary code via unsafe reflection. It’s being leveraged together with vulnerabilities in N-able N-central to deploy cryptocurrency miners. One documented attack saw it exploited in Atlassian Confluence Data Center, chaining shell commands via Axis’s XML-RPC/SOAP APIs. Attackers accessed credentials, parsed documents for sensitive data, and leveraged the flaw to map internal services. Confluence instances with the Remote API feature enabled are particularly at risk. Since Axis 1.x is past end of life, migration or applying a specific patch from July 2026 is advised.
  • REDCap – CVE-2026-90817 (CVSS 9.8): A code injection flaw in the survey-import feature allows arbitrary code execution via crafted HTTP requests. Attackers exploit public survey hashes to trigger the vulnerability—once in, they establish web shells, load in-memory miners (disguised via image file wrappers), and maintain persistence through cron jobs and SSH backdoors.
  • LiteLLM – CVE-2026-40217 (CVSS 8.8): This vulnerability exploits an alternate unprotected channel in LiteLLM to allow remote attackers to rewrite bytecode via a URI endpoint. It’s being used in LLMjacking attacks—campaigns that take over machine learning models or workloads—to escalate privileges and obtain administrative access in sectors such as finance, retail, and transportation worldwide.
  • Synology DSM – CVE-2024-10441 (CVSS 9.8): Due to improper output encoding/escaping, remote attackers can execute code via unspecified vectors. It’s already tied to real ransomware attacks targeting Synology NAS devices, with C3RB3R ransomware found exploiting the flaw.

Urgent Mitigation Deadlines Set

Federal agencies are being directed to patch the Axis, REDCap, and Synology flaws by November 21, 2026, while the LiteLLM issue must be addressed by December 2, 2026. These actions follow confirmation of active exploitation across multiple sectors and geographies.

This move highlights growing adversary interest in targeting tools used for AI, collaboration, and storage. Unsafe reflection, code injection, and unprotected inference channels are becoming go-to methods for attackers aiming at cryptomining, data exfiltration, or LLM misuse. Organizations relying on Synology NAS, running LiteLLM or Axis-dependent services, or using public survey features in REDCap are at immediate risk.

What this means going forward: The inclusion of these flaws in the KEV catalog underscores how rapidly attackers adapt to exploit exposed surface in AI-driven and legacy systems. Monitoring SOAP APIs, managing public survey exposure, and securing AI inference paths are no longer theoretical—but essential. Expect more advisory action from CISA and other regulators as these attack vectors evolve.