ENHANCING ROBUSTNESS AND FAIRNESS IN FEDERATED MACHINE LEARNING FOR PRIVACY-PRESERVING PREDICTIVE HEALTHCARE

Abstract

Federated machine learning (FML) has revolutionized privacy-preserving healthcare by enabling collaborative model development without centralized access to sensitive patient data. Despite its promise, FML faces critical challenges related to robustness and fairness, particularly in environments with heterogeneous data distributions and potential adversarial threats. This paper comprehensively reviews the key robustness and fairness concerns specific to FML in healthcare applications. We analyze recent advancements aimed at mitigating these issues and highlight persistent gaps in current methodologies. Through comparative evaluations using bean plots and performance tables across diverse datasets and techniques, we offer a detailed synthesis of the field. Finally, we propose future research directions to strengthen the reliability and equity of FML systems for healthcare deployment.