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 2209.14987cs.LGcs.CR
收藏

没有免费的午餐在“隐私免费:数据集压缩如何帮助隐私”

No Free Lunch in "Privacy for Free: How does Dataset Condensation Help Privacy"

Google(谷歌);Apple(苹果公司)

Nicholas Carlini, Vitaly Feldman, Milad Nasr

摘要:

旨在保护数据隐私的新方法需要仔细审查。保护隐私的失败很难被发现,但当一个实现“隐私保护”方法的系统受到攻击时,可能会导致灾难性的结果。最近的一项工作入选了ICML 2022年的杰出论文奖(Dong等人,2022)声称,数据集压缩(DC)在训练机器学习模型时显着提高了数据隐私。这一主张得到了特定数据集压缩技术的理论分析和对一些现有成员推断攻击的抵抗性的经验评估的支持。在这篇文章中,我们研究了Dong等人(2022)的工作中的主张,并描述了该方法及其理论分析的实证评估中的主要缺陷。这些缺陷意味着他们的工作没有提供统计学上显著的证据,证明DC在朴素的基线上提高了训练ML模型的隐私性。此外,先前发表的结果表明,DP-SGD是隐私保护ML的标准方法,同时提供了更好的准确性,并实现了(可证明的)较低的成员攻击成功率。

英文摘要:

New methods designed to preserve data privacy require careful scrutiny. Failure to preserve privacy is hard to detect, and yet can lead to catastrophic results when a system implementing a ``privacy-preserving'' method is attacked. A recent work selected for an Outstanding Paper Award at ICML 2022 (Dong et al., 2022) claims that dataset condensation (DC) significantly improves data privacy when training machine learning models. This claim is supported by theoretical analysis of a specific dataset condensation technique and an empirical evaluation of resistance to some existing membership inference attacks. In this note we examine the claims in the work of Dong et al. (2022) and describe major flaws in the empirical evaluation of the method and its theoretical analysis. These flaws imply that their work does not provide statistically significant evidence that DC improves the privacy of training ML models over a naive baseline. Moreover, previously published results show that DP-SGD, the standard approach to privacy preserving ML, simultaneously gives better accuracy and achieves a (provably) lower membership attack success rate.

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。

↑