EvoSafeHarness: Evolving Model- and Domain-Specific Harnesses for Securing Agents
EvoSafeHarness auto-synthesizes per-model, per-domain safety harnesses, cutting prompt-injection attack success on AgentDojo to 0.0% at 82.8% utility.
EvoSafeHarness is an optimization framework that synthesizes deployable safety harnesses for frozen LLM agents in a target domain, jointly searching natural-language policies and executable code logic guided by model behavior, domain specifications, and adversarial review. On DecodingTrust-Agent it reduces average attack success rate from 45.6% to 10.0% at a 3.3-point utility cost, and on AgentDojo reaches 82.8% utility at 0.0% ASR, twice CaMeL's utility at that operating point. It keeps mean ASR below 20% under adaptive PAIR attacks and transfers unchanged to unseen AgentDyn suites. The analysis finds domain semantics determine required safety relations while model and runtime behavior determine enforcement points.
- Automates harness synthesis instead of one-size-fits-all expert-designed defenses
- AgentDojo: 82.8% utility at 0.0% attack success rate, double CaMeL's utility
- DecodingTrust-Agent: average ASR cut from 45.6% to 10.0% at 3.3-point utility cost
- Best scores on Agent-SafetyBench for every tested victim model
- Harness transfers unchanged to unseen AgentDyn benchmark suites
From the original article
1 min read at huggingface.co
Large Language Model (LLM) agents are turning language into real-world effects, making safety necessary against both indirect prompt injections and direct harmful requests. System-level safety harnesses add an enforcement layer beyond model-level defenses, but existing harnesses are usually designed once by experts and applied across heterogeneous models and domains.
By Nanxi Li, Yingzi Ma, Yulong Cao · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.