提出红化框架RedAct，保护智能体执行痕迹中的关键技能不被窃取。
RedAct: Redacting Agent Capability Traces for Procedural Skill Protection
- 通过局部化敏感信息并重写痕迹，保留可审计证据。
- 将技能泄露率从44.7%-67.1%降至基线以下，水印检测率达93.6%-100%。
- 适合关注智能体安全与可审计性的研究者和开发者使用。
用户依赖执行痕迹观察智能体行为、诊断故障并确保问责。这些痕迹包含丰富的过程细节，如工具调用、中间决策和错误恢复逻辑，但可能暴露私有技能，使下游方法在无模型权重或技能文件的情况下复现关键公式、阈值与策略。为量化风险并评估防护效果，我们构建了涵盖7个领域、75项长时程任务和154项精选技能的CapTraceBench基准。同时提出RedAct框架：本地化保护关键信息，重写痕迹以保留验证必要证据，并嵌入行为水印用于溯源分析。在代表性痕迹复用方法下，RedAct将标准化技能迁移（NST）从原始痕迹的44.7%-67.1%降至低于无技能基线水平，同时保留审计证据。其独立行为水印真检测率达93.6%-100%，误报率不超过1.9%。结果表明，公开智能体痕迹应被视为安全接口，选择性红化可在不移除审计证据的前提下降低过程能力泄露风险。
原文摘要 · Abstract (English)
Users rely on execution traces to observe agent behavior, diagnose failures, and ensure accountability. These traces contain rich procedural detail, including tool invocations, intermediate decisions, and error-recovery logic. Yet this detail can expose private procedural skills, allowing downstream methods to recover key formulas, thresholds, and strategies without access to model weights or skill files. To quantify this risk and evaluate protection, we construct CapTraceBench, a benchmark of 75 specialized long-horizon tasks and 154 curated skills across seven domains. We also introduce RedAct, a protected trace release framework that localizes protected key information, rewrites traces while preserving verifier-critical evidence, and embeds behavioral watermarks for downstream provenance analysis. Across representative trace reuse methods, RedAct reduces normalized skill transfer (NST) from 44.7-67.1% on raw traces to below the no-skill baseline, while preserving audit evidence. Its standalone behavioral watermarks reach 93.6-100.0% true detection with a false alarm rate of at most 1.9%. These results frame public agent traces as security interfaces and show that selective redaction can reduce procedural capability leakage without removing audit evidence.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。