Robustness and Fairness Challenges in Federated Machine Learning Models for Privacy-Preserving Predictive Healthcare Applications

Abstract

Federated machine learning (FML) has emerged as a transformative approach for privacy-preserving healthcare applications by enabling collaborative model training without centralizing sensitive patient data. However, significant challenges remain regarding the robustness and fairness of these models, especially in the presence of heterogeneous data distributions and adversarial threats. This paper reviews key robustness and fairness issues faced by FML systems in healthcare contexts, explores state-of-the-art solutions, and suggests future research directions. A literature review synthesizes studies that have addressed these problems. We also provide comparative analyses, including bean plots and performance tables, to elucidate challenges across datasets and methods.