Ditto: Fair and Robust Federated Learning Through Personalization
This work is significant for federated learning practitioners and researchers, as it tackles the critical and often conflicting issues of fairness and robustness in heterogeneous networks.
This paper addresses the competing constraints of fairness and robustness in federated learning by proposing Ditto, a personalized federated learning framework. Ditto achieves competitive performance compared to recent personalization methods and produces more accurate, robust, and fair models than state-of-the-art fair or robust baselines.
Fairness and robustness are two important concerns for federated learning systems. In this work, we identify that robustness to data and model poisoning attacks and fairness, measured as the uniformity of performance across devices, are competing constraints in statistically heterogeneous networks. To address these constraints, we propose employing a simple, general framework for personalized federated learning, Ditto, that can inherently provide fairness and robustness benefits, and develop a scalable solver for it. Theoretically, we analyze the ability of Ditto to achieve fairness and robustness simultaneously on a class of linear problems. Empirically, across a suite of federated datasets, we show that Ditto not only achieves competitive performance relative to recent personalization methods, but also enables more accurate, robust, and fair models relative to state-of-the-art fair or robust baselines.