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 2604.13630cs.CRcs.AI
收藏

SafeHarness:面向LLM代理部署的生命周期集成安全架构

SafeHarness: Lifecycle-Integrated Security Architecture for LLM-based Agent Deployment

Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);Academy of Mathematics and Systems Science, Chinese Academy of Sciences(中国科学院数学与系统科学研究院);Institute of Applied Physics and Computational Mathematics(应用物理与计算数学研究所);Peking University(北京大学)

Xixun Lin, Yang Liu, Yancheng Chen, Yongxuan Wu, Yucheng Ning, Yilong Liu, Nan Sun, Shun Zhang, Bin Chong, Chuan Zhou, Yanan Cao

AI总结:

本文提出SafeHarness架构,通过四个防御层集成到代理生命周期中,解决现有安全方法的结构性不匹配问题,降低不安全行为和攻击成功率。

摘要:

大型语言模型(LLM)代理的性能严重依赖执行 harness,即协调工具使用、上下文管理和状态持久化的系统层。然而,这一关键架构也使其成为高价值的攻击面:harness 层的单点 compromise 可能导致整个执行流程的崩溃。本文引入 SafeHarness,一种安全架构,其中四个提出的防御层直接集成到代理生命周期中,以解决上述显著限制:对抗性上下文过滤在输入处理阶段,分层因果验证在决策阶段,权限分离的工具控制在行动执行阶段,以及安全回滚与自适应退化在状态更新阶段。所提出的跨层机制将这些层连接起来,当检测到持续异常时,提高验证严格性,触发回滚,并收紧工具权限。我们评估了 SafeHarness 在跨多样化的 harness 配置的基准数据集上,与四种安全基线在五个攻击场景中进行比较,涵盖六个威胁类别。与未保护基线相比,SafeHarness 实现了平均约 38% 的 UBR 减少和 42% 的 ASR 减少,显著降低了不安全行为率和攻击成功率,同时保持核心任务的实用性。

英文摘要:

The performance of large language model (LLM) agents depends critically on the execution harness, the system layer that orchestrates tool use, context management, and state persistence. Yet this same architectural centrality makes the harness a high-value attack surface: a single compromise at the harness level can cascade through the entire execution pipeline. We observe that existing security approaches suffer from structural mismatch, leaving them blind to harness-internal state and unable to coordinate across the different phases of agent operation. In this paper, we introduce \safeharness{}, a security architecture in which four proposed defense layers are woven directly into the agent lifecycle to address above significant limitations: adversarial context filtering at input processing, tiered causal verification at decision making, privilege-separated tool control at action execution, and safe rollback with adaptive degradation at state update. The proposed cross-layer mechanisms tie these layers together, escalating verification rigor, triggering rollbacks, and tightening tool privileges whenever sustained anomalies are detected. We evaluate \safeharness{} on benchmark datasets across diverse harness configurations, comparing against four security baselines under five attack scenarios spanning six threat categories. Compared to the unprotected baseline, \safeharness{} achieves an average reduction of approximately 38\% in UBR and 42\% in ASR, substantially lowering both the unsafe behavior rate and the attack success rate while preserving core task utility.

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。

↑