Decentralized Federated Learning for Private Smart Healthcare: A Survey †

Abstract

This research explores the use of decentralized federated learning (DFL) in healthcare, focusing on overcoming the shortcomings of traditional centralized FL systems. DFL is proposed as a solution to enhance data privacy and improve system reliability by reducing dependence on central servers and increasing local data control. The research adopts a systematic literature review, following PRISMA guidelines, to provide a comprehensive understanding of DFL’s current applications and challenges within healthcare. The review synthesizes findings from various sources to identify the benefits and gaps in existing research, proposing research questions to further investigate the feasibility and optimization of DFL in medical environments. The study identifies four key challenges for DFL: security and privacy, communication efficiency, data and model heterogeneity, and incentive mechanisms. It discusses potential solutions, such as advanced cryptographic methods, optimized communication strategies, adaptive learning models, and robust incentive frameworks, to address these challenges. Furthermore, the research highlights the potential of DFL in enabling personalized healthcare through large, distributed data sets across multiple medical institutions. This study fills a critical gap in the literature by systematically reviewing DFL technologies in healthcare, offering valuable insights into applications, challenges, and future research directions that could improve the security, efficiency, and equity of healthcare data management.