FARMA: Falsifying Agent Decision Logs
Inevitable_Fee1895 · reddit · 2026-07-12
This paper introduces FARMA, a novel memory poisoning attack targeting an agent's own decision logs/reasoning traces rather than retrieved knowledge base facts. The attack operates in two steps: injecting seemingly benign log seeds, and then continuously amplifying these fake logs as "precedents" to trick the agent into treating skipped checks as established practice.
Tested on an EHR agent using GPT-4o-mini, GPT-4o, and Llama 3.3 70B, the attack achieved a 100% success rate without defenses, and maintained 100% against keyword filtering and A-MemGuard. The proposed defense, SENTINEL, uses structured analysis of reasoning trajectories to reduce the success rate to 0% with zero false positives across 326 normal trajectories. However, the authors acknowledge that if attackers know its specific heuristic rules, the defense's effectiveness will be weakened.
More from Safety
- Hugging Face chief says U.S. guardrails forced a Chinese model into a real cyber defense — Nunki08 · 2026-07-21
- AgentBaiting uses 600 fake MCP and Skills listings to lure AI assistants — TechNadu · 2026-07-21
- Enterprise LLM security course focuses on protecting agentic AI apps — Independentgoats · 2026-07-21
- YouTube is cracking down on mass-produced synthetic videos, users say — No_Link7744 · 2026-07-21
- Suno breach talk is being muted in Discord, Reddit users say — chuckbeefcake · 2026-07-21
- Native and Cyera link data discovery to cloud access controls for AI use — TechNadu · 2026-07-21