CuraSec

tag: Ai-Ml-Security · 2 items

Act CISA KEV: Critical MLflow vulnerability under active exploitation
2026-08-20 · BleepingComputer · source ↗ #cisa-kev#mlflow#ai-ml-security
  • Engineer — Act: CISA’s active-exploitation warning effectively signals KEV listing; teams running MLflow in AI/ML pipelines should patch to the fixed version immediately and audit pipeline access logs for signs of prior compromise.
  • SOC/IR — Plan: The item confirms active exploitation but provides no IOCs or TTPs to hunt on; pull the full CISA advisory for indicators, then build detection rules targeting anomalous MLflow API or model-registry access patterns.
  • Leader — Plan: Confirm whether data science or engineering teams operate MLflow, then verify patching is tracked to completion — CISA exploitation warnings on AI/ML tooling are increasingly likely to surface in customer security questionnaires.
Act CISA KEV: Actively Exploited RCE in Ray AI/ML Framework
2026-08-18 · The Hacker News · source ↗ #cisa-kev#rce#ai-ml-security
  • Engineer — Act: CISA KEV listing confirms active exploitation of a critical RCE in Ray, a widely-used Python distributed computing framework for AI/ML workloads; patch Ray immediately and, if patching is delayed, restrict external access to Ray dashboard and cluster endpoints.
  • SOC/IR — Act: Active exploitation confirmed via KEV; hunt for unauthorized code execution originating from Ray cluster nodes and sweep for internet-exposed Ray dashboards in your environment, prioritizing ML infrastructure that may not be covered by standard EDR.
  • Leader — Plan: If your organization runs AI/ML workloads, ask engineering to confirm whether Ray is deployed and to report patch status; KEV listing makes this likely to surface in auditor or customer questionnaires about your ML infrastructure security posture.