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 2109.11429cs.LGcs.AIcs.CRcs.CY
收藏

罗宾汉和马太效应:差分隐私对合成数据的影响不同

Robin Hood and Matthew Effects: Differential Privacy Has Disparate Impact on Synthetic Data

University College London(伦敦大学学院);Hazy

Georgi Ganev, Bristena Oprisanu, Emiliano De Cristofaro

摘要:

使用差分隐私(DP)训练的生成模型可用于生成合成数据,同时最小化隐私风险。我们分析了DP对这些模型的影响,具体而言,研究:1)合成数据中类/子组的大小,以及2)在它们上运行的分类任务的准确性。我们还评估了各种程度的不平衡和隐私预算的影响。我们的分析使用三种最先进的DP模型(PrivBayes,DP-WGAN和PATE-GAN),并显示DP在生成的合成数据中产生相反的大小分布。它影响到多数和少数阶层/亚群体之间的差距;在某些情况下,缩小差距(“罗宾汉”效应),在另一些情况下,扩大差距(“马太”效应)。无论哪种方式,这都会对合成数据上的分类任务的准确性产生(类似的)不同的影响,对数据中代表性不足的子部分的影响更大。因此,当在合成数据上训练模型时,可能会产生不均匀地对待不同亚群的风险,从而导致不可靠或不公平的结论。

英文摘要:

Generative models trained with Differential Privacy (DP) can be used to generate synthetic data while minimizing privacy risks. We analyze the impact of DP on these models vis-a-vis underrepresented classes/subgroups of data, specifically, studying: 1) the size of classes/subgroups in the synthetic data, and 2) the accuracy of classification tasks run on them. We also evaluate the effect of various levels of imbalance and privacy budgets. Our analysis uses three state-of-the-art DP models (PrivBayes, DP-WGAN, and PATE-GAN) and shows that DP yields opposite size distributions in the generated synthetic data. It affects the gap between the majority and minority classes/subgroups; in some cases by reducing it (a "Robin Hood" effect) and, in others, by increasing it (a "Matthew" effect). Either way, this leads to (similar) disparate impacts on the accuracy of classification tasks on the synthetic data, affecting disproportionately more the underrepresented subparts of the data. Consequently, when training models on synthetic data, one might incur the risk of treating different subpopulations unevenly, leading to unreliable or unfair conclusions.

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。

↑