arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.14271cs.CLcs.CY
收藏

审查代理安全

Auditing Agent Harness Safety

UCB(加州大学伯克利分校);WISC(威斯康星大学);STANFORD(斯坦福大学);MSR1(微软研究院)

Chengzhi Liu, Yichen Guo, Yepeng Liu, Yuzhe Yang, Qianqi Yan, Xuandong Zhao, Wenyue Hua, Sheng Liu, Sharon Li, Yuheng Bu, Xin Eric Wang

AI总结:

本文提出HarnessAudit框架,用于审查执行轨迹中的边界合规性、执行保真度和系统稳定性,揭示多代理Harness中的安全风险,并通过HarnessAudit-Bench验证了安全风险与轨迹长度、领域和代理角色的关系。

摘要:

LLM代理越来越多地在执行Harness中运行,这些Harness负责调度工具、分配资源并路由消息。然而,Harness可能在轨迹中访问未授权资源或泄露上下文给错误的代理,但输出级评估无法检测这些失败。本文提出HarnessAudit框架,用于审查完整执行轨迹的边界合规性、执行保真度和系统稳定性,并通过HarnessAudit-Bench验证了安全风险与轨迹长度、领域和代理角色的关系。

英文摘要:

LLM agents increasingly run inside execution harnesses that dispatch tools, allocate resources, and route messages between specialized components. However, a harness can return a correct, benign answer over a trajectory that accesses unauthorized resources or leaks context to the wrong agent. Output-level evaluation cannot see these failures, yet most safety benchmarks score only final outputs or terminal states, even though many violations occur mid-trajectory rather than at termination. The central question is whether the harness respects user intent, permission boundaries, and information-flow constraints throughout execution. To address this gap, we propose HarnessAudit, a framework that audits full execution trajectories across boundary compliance, execution fidelity, and system stability, with a focus on multi-agent harnesses where these risks are most pronounced. We further introduce HarnessAudit-Bench, a benchmark of 210 tasks across eight real-world domains, instantiated in both single-agent and multi-agent configurations with embedded safety constraints. Evaluating ten harness configurations across frontier models and three multi-agent frameworks, we find that: (i) task completion is misaligned with safe execution, and violations accumulate with trajectory length; (ii) safety risks vary across domains, task types, and agent roles; (iii) most violations concentrate in resource access and inter-agent information transfer; and (iv) multi-agent collaboration expands the safety risk surface, while harness design sets the upper bound of safe deployment.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑