SecureClaw: 夺回对LLM智能体的控制
SecureClaw: Clawing Back Control of LLM Agents
AI总结:
针对工具使用型LLM智能体的双重安全漏洞,提出双边界架构SecureClaw,在效果汇点实施授权、在读边界实施明文隔离,通过预览-提交协议和可信网关实现安全控制,在多个基准上保持可用性的同时将攻击成功率降至接近零。
摘要:
使用工具的大型语言模型(LLM)智能体面临两种不同的安全漏洞:未经授权的外部操作以及在最终输出检查介入之前运行时内部敏感明文的暴露。现有防御通常只保护一个边界(规划器/运行时或动作汇点),因此本身无法同时保护两个表面。我们提出SecureClaw,一种双边界架构,在效果汇点实施授权,在读边界实施明文隔离。敏感读取通过一个可信网关,该网关用不透明句柄替换原始值,在评估部署中,还使用有界摘要作为显式解密接口。改变外部状态的写入遵循PREVIEW→COMMIT协议,其中只有可信执行者才能提交策略授权的确切规范请求。运行时仍然可以基于摘要和符号引用进行规划,但不能直接解引用秘密或执行副作用。在AgentDojo、AgentLeak和Agent Security Bench (ASB)上,SecureClaw是我们在通用测试框架中评估的唯一一种同时保持可用任务效用并在ASB上实现0%攻击成功率(ASR)、在AgentDojo上实现0.64% ASR、在AgentLeak的攻击并行通道上实现3.23%总体泄漏(衡量最终输出和内部中继泄漏)的防御方法。
英文摘要:
Tool-using large language model (LLM) agents face two distinct security failures: unauthorized external actions and exposure of sensitive plaintext inside the runtime before any final output check can intervene. Existing defenses usually protect one boundary, either the planner/runtime or the point where external actions are executed, and therefore do not by themselves secure both surfaces. We present SecureClaw, a dual-boundary architecture that enforces authorization in a trusted executor and keeps raw protected data outside the LLM runtime. Sensitive reads pass through a trusted gateway that replaces raw values with reference IDs. In the evaluated deployment, the gateway also provides short planning summaries that deliberately release selected information, a process called declassification. Writes that change external state follow a PREVIEW$\rightarrow$COMMIT protocol in which only a trusted executor may execute the exact request authorized by policy. The runtime can still plan over summaries and reference IDs, but cannot directly resolve the IDs into secrets or perform side effects. Across the AgentDojo, AgentLeak, and Agent Security Bench (ASB) benchmarks, SecureClaw is the only defense we evaluate in a common harness that simultaneously retains usable task utility and achieves 0% attack success rate (ASR) on ASB, 0.64% ASR on AgentDojo, and 3.23% overall leak on AgentLeak's attacked parity lane, which measures final-output and internal-relay leakage. Code is available at https://github.com/ponyyuhan/SecureClaw.
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。