P2Skill：面向云-本地大语言模型推理系统的隐私保护技能蒸馏

中文摘要 AI 辅助

云-本地大语言模型（LLM）推理系统可利用云端大模型的推理能力，同时保护个人设备上的敏感用户数据。发往云端的请求必须排除个人身份信息（PII），以防止外部数据泄露。现有隐私保护方法依赖提示扰动、实体掩码或模型微调，但这些方法可能扭曲上下文语义，或需要额外训练。本文提出P2Skill，一种基于提示的技能蒸馏方法：本地小型语言模型（SLM）遵循技能提示，自主完成分解、感知PII的路由、 paraphrasing（改写）及重构。云端LLM会根据执行失败迭代优化技能，使本地SLM能泛化至超出记忆的PII模式，因此P2Skill无需隐私特定微调或学习辅助检测器。在四领域基准上的评估显示，P2Skill的隐私保护推理质量较先前基线分别高出1.69倍和3.66倍。

英文摘要

Cloud-local LLM inference systems have the potential to use the reasoning capability of large cloud models while protecting sensitive user data on personal devices. Cloud-bound requests must exclude personally identifiable information (PII) to prevent external data leakage. Existing privacy-preserving methods rely on prompt perturbation, entity masking, or model fine-tuning, but these approaches may distort contextual semantics or require additional training. This paper proposes P2Skill, a prompt-based skill distillation method in which a local small language model (SLM) autonomously performs decomposition, PII-aware routing, paraphrasing, and reconstruction by following the skill prompts. Skills are iteratively refined from execution failures by a cloud LLM, enabling the local SLM to generalize beyond memorized PII patterns, and therefore P2Skill requires no privacy-specific fine-tuning or learned auxiliary detectors. Evaluation on a four-domain benchmark shows that P2Skill achieves $1.69\times$ and $3.66\times$ higher privacy-preserved inference quality than previous baselines.

发表机构

• Korea University（高丽大学）

机构由 AI 辅助整理，请以论文原文为准。