Open-source repo ships detection rules for agentic attack frameworks like ARTEX and Cairn
blaizedsouza · x · 2026-11-22
ThreatDefence released agentic-attack-detections, an open-source repo with detection rules, KQL hunting queries and IOCs (Sigma, Suricata, YARA) for AI agent attack frameworks: ARTEX/AntiFox, Cairn, PentAGI, HexStrike, CyberStrikeAI, Strix, Hermes Agent and Agent Zero — tools that hand intrusions to AI agents which plan, run real attack tools and persist until the goal is met. Rules are built from source-code review and public reporting; shared behaviors (agents running tools with approval prompts off, servers calling Chinese model APIs, tool-call bursts, rotating-proxy probing) get their own rules so one rule covers every framework.
More from Safety
- MIRI's old alignment strategy under scrutiny as community flags lab conflicts of interest — jessi_cata · 2026-11-22
- AI safety debate: Geoffrey Irving floats short reversible unilateral pauses as coalition-building probes — catherineols · 2026-11-22
- Cluaiz adds OS sandboxing to its AI agent after three real-machine failures — CommissionOdd3082 · 2026-11-22
- David Friedberg: Restricting Open Source AI Is a Path to Tyranny — kevinnbass · 2026-11-22
- Hacker tricked Meta's Muse agent into silently stealing encrypted passwords — AccBalanced · 2026-11-22
- NIH R35 grant scored 18 still unfunded as scientist details months of silence — anshulkundaje · 2026-11-22