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 2512.08809cs.CRcs.AIcs.LG
收藏

PrivTune: 通过设备-云协作实现高效且隐私保护的大型语言模型微调

PrivTune: Efficient and Privacy-Preserving Fine-Tuning of Large Language Models via Device-Cloud Collaboration

City University of Hong Kong(香港城市大学);City University of Hong Kong (Dongguan)(香港城市大学(东莞));University of Melbourne(墨尔本大学)

Yi Liu, Weixiang Han, Chengjun Cai, Xingliang Yuan, Cong Wang

AI总结:

PrivTune通过分割学习框架,在设备-云协作中实现高效且隐私保护的大型语言模型微调,有效降低攻击成功率并保持性能。

摘要:

随着大型语言模型的兴起,服务提供商提供语言模型作为服务,使用户能够通过上传私有数据集来微调定制模型。然而,这引发了敏感数据泄露的担忧。以往的方法依赖于设备-云协作框架中的差分隐私,难以在隐私和实用性之间取得平衡,使用户面临推断攻击或降低微调性能的风险。为了解决这一问题,我们提出了PrivTune,一种通过分割学习(SL)实现的高效且隐私保护的微调框架。PrivTune的核心思想是将精心设计的噪声注入到SL底层模型的token表示中,使每个token类似于n-hop间接邻居。PrivTune将这一过程建模为一个优化问题,以计算最优噪声向量,以符合防御-实用性目标。在此基础上,它调整参数(即均值)的d_χ-隐私噪声分布,以与优化方向对齐,并根据token的重要性缩放噪声,以最小化失真。在五个数据集(涵盖分类和生成任务)上,针对三种嵌入反向和三种属性推断攻击的实验表明,使用RoBERTa在斯坦福情感树库数据集上,PrivTune将攻击成功率降至10%,仅导致实用性性能下降3.33%,优于最先进的基线方法。

英文摘要:

With the rise of large language models, service providers offer language models as a service, enabling users to fine-tune customized models via uploaded private datasets. However, this raises concerns about sensitive data leakage. Prior methods, relying on differential privacy within device-cloud collaboration frameworks, struggle to balance privacy and utility, exposing users to inference attacks or degrading fine-tuning performance. To address this, we propose PrivTune, an efficient and privacy-preserving fine-tuning framework via Split Learning (SL). The key idea of PrivTune is to inject crafted noise into token representations from the SL bottom model, making each token resemble the $n$-hop indirect neighbors. PrivTune formulates this as an optimization problem to compute the optimal noise vector, aligning with defense-utility goals. On this basis, it then adjusts the parameters (i.e., mean) of the $d_χ$-Privacy noise distribution to align with the optimization direction and scales the noise according to token importance to minimize distortion. Experiments on five datasets (covering both classification and generation tasks) against three embedding inversion and three attribute inference attacks show that, using RoBERTa on the Stanford Sentiment Treebank dataset, PrivTune reduces the attack success rate to 10% with only a 3.33% drop in utility performance, outperforming state-of-the-art baselines.

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。

↑