FairDP: Certified Fairness with Differential Privacy
FairDP: Certified Fairness with Differential Privacy
复制标题
DOI：
10.48550/arxiv.2305.16474
复制
发表时间：
2023-05
期刊：
影响因子：
--
通讯作者：
K. Tran;Ferdinando Fioretto;Issa Khalil;M. Thai;Nhathai Phan
中科院分区：
文献类型：
--
作者：
K. Tran;Ferdinando Fioretto;Issa Khalil;M. Thai;Nhathai Phan
本文介绍了 FairDP，这是一种旨在通过差分隐私 (DP) 实现认证公平性的新颖机制。 FairDP 为不同的个体群体独立训练模型，使用特定于群体的剪裁术语来评估和限制 DP 的不同影响。在整个训练过程中，该机制逐步整合群体模型中的知识，形成一个在下游任务中平衡隐私、实用性和公平性的综合模型。广泛的理论和实证分析验证了 FairDP 的有效性，并与现有方法相比改进了模型效用、隐私和公平性之间的权衡。
This paper introduces FairDP, a novel mechanism designed to achieve certified fairness with differential privacy (DP). FairDP independently trains models for distinct individual groups, using group-specific clipping terms to assess and bound the disparate impacts of DP. Throughout the training process, the mechanism progressively integrates knowledge from group models to formulate a comprehensive model that balances privacy, utility, and fairness in downstream tasks. Extensive theoretical and empirical analyses validate the efficacy of FairDP and improved trade-offs between model utility, privacy, and fairness compared with existing methods.