AFed: Algorithmic Fair Federated Learning

2025-01-06

AI Summary

To address the lack of group fairness in federated learning caused by heterogeneous client data distributions, this paper proposes AFed—a privacy-preserving debiasing framework that requires no access to raw client data. Methodologically, AFed introduces a dual-path architecture comprising a server-side conditional generator (AFed-G) and client-side conditional GANs (AFed-GAN), enabling global data distribution modeling under strict privacy constraints while jointly optimizing fairness and model performance. The approach integrates conditional generation, adversarial training, and theoretical fairness analysis to provably satisfy statistical fairness criteria—including Demographic Parity and Equalized Odds. Extensive experiments across multiple real-world benchmarks demonstrate that AFed achieves an average 35% improvement in group fairness metrics with negligible accuracy degradation, significantly outperforming existing fair federated learning baselines.

Technology Category

Machine Learning: Distributed Machine Learning & Federated LearningComputer Vision: Bias, Fairness & PrivacyNatural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

User Modeling, Personalization and Recommendation: Federated recommendation systems and personalizationSocial Networks and Social Media: Fairness and bias in social network and social media analysisSearch and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Federated Learning (FL) has gained significant attention as it facilitates collaborative machine learning among multiple clients without centralizing their data on a server. FL ensures the privacy of participating clients by locally storing their data, which creates new challenges in fairness. Traditional debiasing methods assume centralized access to sensitive information, rendering them impractical for the FL setting. Additionally, FL is more susceptible to fairness issues than centralized machine learning due to the diverse client data sources that may be associated with group information. Therefore, training a fair model in FL without access to client local data is important and challenging. This paper presents AFed, a straightforward yet effective framework for promoting group fairness in FL. The core idea is to circumvent restricted data access by learning the global data distribution. This paper proposes two approaches: AFed-G, which uses a conditional generator trained on the server side, and AFed-GAN, which improves upon AFed-G by training a conditional GAN on the client side. We augment the client data with the generated samples to help remove bias. Our theoretical analysis justifies the proposed methods, and empirical results on multiple real-world datasets demonstrate a substantial improvement in AFed over several baselines.

Problem

• Federated Learning
• Data Fairness
• Distributed Data

Innovation

• AFed
• Federated Learning
• Bias Mitigation