arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2311.04044cs.CLcs.CR
收藏

PrivLM-Bench:面向语言模型的多层级隐私评估基准

PrivLM-Bench: A Multi-level Privacy Evaluation Benchmark for Language Models

The Hong Kong University of Science and Technology(香港科技大学);Center for Data Science, AAIS, Peking University(北京大学AAIS数据科学中心);Amazon.com Inc(亚马逊公司)

Haoran Li, Dadi Guo, Donghao Li, Wei Fan, Qi Hu, Xin Liu, Chunkit Chan, Duanyi Yao, Yuan Yao, Yangqiu Song

AI总结:

本文提出多视角隐私评估基准PrivLM-Bench,通过定义隐私目标、构建隐私微调统一流程并执行隐私攻击,实证量化并公平比较了主流语言模型的隐私泄露风险。

摘要:

语言模型的快速发展为模型和用户带来了前所未有的可及性与使用量。一方面,强大的语言模型在众多下游NLP任务上取得了最先进的性能。另一方面,不受限制的模型访问可能带来数据泄露的恶意隐私风险,这受到了越来越多的关注。为了解决这些问题,许多近期的研究提出了具有差分隐私(DP)的隐私保护语言模型(PPLMs)。遗憾的是,不同的DP实现使得在现有PPLMs之间进行公平比较面临挑战。在本文中,我们提出了PrivLM-Bench,这是一个多视角的隐私评估基准,旨在直观且实证地量化语言模型的隐私泄露。PrivLM-Bench不再仅仅报告DP参数,而是揭示了实际使用过程中被忽视的推理数据隐私。PrivLM-Bench首先明确定义了多方面的隐私目标。随后,PrivLM-Bench构建了统一的流程来执行隐私微调。最后,PrivLM-Bench以预定义的隐私目标在语言模型上执行现有的隐私攻击,以此作为实证评估结果。这些实证攻击结果被用于公平、直观地评估各种PPLMs的隐私泄露情况。我们在GLUE的三个数据集上对主流语言模型进行了广泛实验。

英文摘要:

The rapid development of language models (LMs) brings unprecedented accessibility and usage for both models and users. On the one hand, powerful LMs achieve state-of-the-art performance over numerous downstream NLP tasks. On the other hand, more and more attention is paid to unrestricted model accesses that may bring malicious privacy risks of data leakage. To address these issues, many recent works propose privacy-preserving language models (PPLMs) with differential privacy (DP). Unfortunately, different DP implementations make it challenging for a fair comparison among existing PPLMs. In this paper, we present PrivLM-Bench, a multi-perspective privacy evaluation benchmark to empirically and intuitively quantify the privacy leakage of LMs. Instead of only reporting DP parameters, PrivLM-Bench sheds light on the neglected inference data privacy during actual usage. PrivLM-Bench first clearly defines multi-faceted privacy objectives. Then, PrivLM-Bench constructs a unified pipeline to perform private fine-tuning. Lastly, PrivLM-Bench performs existing privacy attacks on LMs with pre-defined privacy objectives as the empirical evaluation results. The empirical attack results are used to fairly and intuitively evaluate the privacy leakage of various PPLMs. We conduct extensive experiments on three datasets of GLUE for mainstream LMs.

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。

↑