Screwtape:用于监控红队测试的对抗性 transcript 编辑
Screwtape: Adversarial Transcript Editing for Monitor Red-Teaming
AI总结:
该研究提出Screwtape流水线,将真实智能体transcript转为规避仅动作监控器的攻击,在94个Claude Code和Codex transcript上测试,成功生成35次规避攻击,可用于修复监控器盲点。
摘要:
我们提出了Screwtape,一种将真实智能体 transcript 转换为可规避仅动作监控器的攻击的流水线。一个LLM攻击者编辑 transcript,使智能体引发指定的危害,例如泄露API密钥或生成未受监控的智能体。由于编辑后的动作从未被执行,我们使用一个LLM验证器来拒绝与原始 transcript 不一致或低于严重程度阈值的攻击。攻击者会修改编辑内容,直到监控器不再标记它。我们将Screwtape应用于Redwood Research员工的94个Claude Code和Codex transcript,针对我们自己的监控器进行优化。在252次运行中,它产生了35次通过验证器并规避这些监控器的攻击。其中6次经人工审查确认为有效,还规避了前沿开发者提供的保留监控器。因此,我们证明,无法访问开发者基础设施的红队测试人员可以找到开发者监控器的已确认故障。我们预计,当该流水线在开发者基础设施内部运行、使用缺乏拒绝保障措施的前沿攻击者模型并直接针对生产监控器优化时,会发现更多攻击。开发者可以利用这些攻击修复监控器的盲点,评估人员可以将其作为关于监控器安全主张的反例。
英文摘要:
We present Screwtape, a pipeline that converts real agent transcripts into attacks that evade action-only monitors. An LLM attacker edits a transcript so that the agent causes a specified harm, such as leaking an API key or spawning an unmonitored agent. The edited actions are never executed, so we use an LLM verifier to reject attacks that are inconsistent with the original transcript or below a severity bar. The attacker revises the edit until the monitor no longer flags it. We applied Screwtape to 94 Claude Code and Codex transcripts from Redwood Research staff, optimizing against our own monitors. Across 252 runs, it produced 35 attacks that passed the verifier and evaded these monitors. Six of them, confirmed valid by human review, also evaded a held-out monitor provided by a frontier developer. We therefore demonstrate that a red-teamer without access to a developer's infrastructure can find confirmed failures of the developer's monitor. We expect to find more attacks when the pipeline runs inside the developer's infrastructure, with frontier attacker models that lack refusal safeguards and direct optimization against the production monitor. Developers can use these attacks to fix monitor blind spots, and evaluators can use them as counterexamples to safety claims about the monitor.
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。