This website requires JavaScript.
DOI: 10.1145/3543873.3587577

Fairness-aware Differentially Private Collaborative Filtering

Zhenhuan YangYingqiang GeCongzhe SuDingxian WangXiaoting ZhaoYiming Ying
Mar 2023
Recently, there has been an increasing adoption of differential privacyguided algorithms for privacy-preserving machine learning tasks. However, theuse of such algorithms comes with trade-offs in terms of algorithmic fairness,which has been widely acknowledged. Specifically, we have empirically observedthat the classical collaborative filtering method, trained by differentiallyprivate stochastic gradient descent (DP-SGD), results in a disparate impact onuser groups with respect to different user engagement levels. This, in turn,causes the original unfair model to become even more biased against inactiveusers. To address the above issues, we propose \textbf{DP-Fair}, a two-stageframework for collaborative filtering based algorithms. Specifically, itcombines differential privacy mechanisms with fairness constraints to protectuser privacy while ensuring fair recommendations. The experimental results,based on Amazon datasets, and user history logs collected from Etsy, one of thelargest e-commerce platforms, demonstrate that our proposed method exhibitssuperior performance in terms of both overall accuracy and user group fairnesson both shallow and deep recommendation models compared to vanilla DP-SGD.