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 2605.08646cs.LGcs.CLcs.DC
收藏

PAAC:隐私感知的代理设备-云协作

PAAC: Privacy-Aware Agentic Device-Cloud Collaboration

Purdue University(普渡大学);University of Exeter(埃克塞特大学)

Liangqi Yuan, Wenzhi Fang, Shiqiang Wang, Christopher G. Brinton

AI总结:

PAAC框架通过设备-云边界对齐规划-执行分解,使角色专业化成为隐私机制,提升隐私与准确性的平衡,平均准确率提升15-36%,泄漏减少2-6倍。

摘要:

大型语言模型(LLM)代理面临结构张力:云代理提供强推理但暴露用户数据,而设备端代理保护隐私却牺牲整体能力。现有设备-云设计将此边界视为计算划分而非适合代理工作负载的信任边界,现有净化器迫使在策略灵活性和结构忠实性工具调用之间做出选择。在本工作中,我们开发了PAAC,一个隐私感知的代理框架,通过将规划-执行分解与设备-云边界对齐,使角色专业化本身成为隐私机制。云代理在类型占位符令牌上推理,保留每个敏感值的推理角色而丢弃其内容,而设备端代理识别敏感跨度并将其每一步的执行结果提炼为紧凑的关键发现。净化器将设备端LLM限制在提出需掩码的跨度,而确定性注册表执行所有替换和反转,保持动作可直接在设备上执行。在三个严格隐私设置下的代理基准测试中,PAAC主导隐私与准确性的帕累托前沿,平均准确率提升15-36%,泄漏减少2-6倍,隐私目标在非固定实体分类法外的边际最大。我们还在17个额外的基准测试中发现一致改进,涵盖10个领域,包括数学、科学和金融。

英文摘要:

Large language model (LLM) agents face a structural tension: cloud agents provide strong reasoning but expose user data, while on-device agents preserve privacy at the cost of overall capability. Existing device-cloud designs treat this boundary as a compute split rather than a trust boundary suited to agentic workloads, and existing sanitizers force a choice between policy flexibility and the structural fidelity tool calls require. In this work, we develop PAAC, a privacy-aware agentic framework that aligns planner--executor decomposition with the device-cloud boundary so that role specialization itself becomes the privacy mechanism. The cloud agent reasons over typed placeholder tokens that preserve each sensitive value's reasoning role while discarding its content, while the on-device agent identifies sensitive spans and distills each step's execution outcome into compact key findings. Sanitization confines the on-device LLM to proposing which spans to mask, while a deterministic registry performs all substitution and reversal, keeping actions directly executable on device. On three agentic benchmarks under strict privacy settings, PAAC dominates the Pareto frontier of privacy and accuracy, improving average accuracy by 15-36\% and reducing average leakage by 2-6$\times$ over state-of-the-art device-cloud baselines, with the largest margins on privacy targets outside fixed entity taxonomies. We find consistent improvements on 17 additional benchmarks spanning 10 domains, including math, science, and finance.

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。

↑