Federated Learning at the Forefront of Fairness: A Multifaceted Perspective

TL;DR

This survey addresses fairness in Federated Learning amid heterogeneous client constraints by proposing a multifaceted taxonomy that separates model-performance-oriented and capability-oriented fairness. It comprehensively reviews notions of fairness, a broad set of fairness-aware approaches (optimization, aggregation, personalization, resource allocation, and incentives), and evaluation metrics, illustrating how to balance equity with model performance. The paper highlights practical algorithms (e.g., sharpness-aware optimization, Lyapunov-based client selection, Shapley-value personalization) and discusses open directions, including fairness-accuracy/privacy-generalization-utility trade-offs. Its contributions provide a structured foundation for researchers to design fair, robust FL systems with scalable evaluation and targeted future work.

Abstract

Fairness in Federated Learning (FL) is emerging as a critical factor driven by heterogeneous clients' constraints and balanced model performance across various scenarios. In this survey, we delineate a comprehensive classification of the state-of-the-art fairness-aware approaches from a multifaceted perspective, i.e., model performance-oriented and capability-oriented. Moreover, we provide a framework to categorize and address various fairness concerns and associated technical aspects, examining their effectiveness in balancing equity and performance within FL frameworks. We further examine several significant evaluation metrics leveraged to measure fairness quantitatively. Finally, we explore exciting open research directions and propose prospective solutions that could drive future advancements in this important area, laying a solid foundation for researchers working toward fairness in FL.

Figures (1)

• Figure 1: A taxonomy of fairness in federated learning, categorizing into model performance-oriented and capability-oriented approaches.