arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2310.13786stat.MLcs.AIcs.LG
收藏

机器学习模型上成员推断攻击的基本极限

Fundamental Limits of Membership Inference Attacks on Machine Learning Models

Université Paris-Saclay, CNRS, Laboratoire de mathématiques d’Orsay(巴黎萨克雷大学,CNRS,奥赛数学实验室)

Eric Aubinais, Elisabeth Gassiat, Pablo Piantanida

AI总结:

本文从理论上刻画了机器学习模型成员推断攻击的基本统计极限,证明过拟合非线性回归中攻击可高概率成功,并发现数据离散化可增强安全性,其界限由数据分布多样性常数决定。

摘要:

成员推断攻击(MIA)能够揭示某个特定数据点是否属于训练数据集,从而可能暴露个人的敏感信息。本文通过探讨机器学习模型上成员推断攻击的统计基本限制,提供了理论保证。更具体地,我们首先推导出控制此类攻击有效性和成功率的统计量。随后,我们从理论上证明,在过拟合学习过程的非线性回归设置中,攻击可能具有很高的成功概率。最后,我们研究了若干情形,并为该统计量提供了界限。有趣的是,我们的发现表明,对数据进行离散化可能增强学习过程的安全性。具体而言,该安全性被证明受一个常数限制,该常数量化了底层数据分布的多样性。我们通过简单的模拟实验展示了这些结果。

英文摘要:

Membership inference attacks (MIA) can reveal whether a particular data point was part of the training dataset, potentially exposing sensitive information about individuals. This article provides theoretical guarantees by exploring the fundamental statistical limitations associated with MIAs on machine learning models at large. More precisely, we first derive the statistical quantity that governs the effectiveness and success of such attacks. We then theoretically prove that in a non-linear regression setting with overfitting learning procedures, attacks may have a high probability of success. Finally, we investigate several situations for which we provide bounds on this quantity of interest. Interestingly, our findings indicate that discretizing the data might enhance the learning procedure's security. Specifically, it is demonstrated to be limited by a constant, which quantifies the diversity of the underlying data distribution. We illustrate those results through simple simulations.

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。

↑