Ditto: Fair and Robust Federated Learning Through Personalization

TL;DR

Ditto introduces a personalized federated learning framework that enhances fairness and robustness across heterogeneous devices, balancing performance and security in federated systems.

Contribution

The paper proposes a scalable personalized federated learning framework, Ditto, that inherently improves fairness and robustness simultaneously in heterogeneous networks.

Findings

Achieves competitive performance with recent personalization methods.

Enables more accurate, robust, and fair models compared to state-of-the-art baselines.

Provides theoretical analysis on fairness and robustness in linear problems.

Abstract

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.