Federated Learning: Healthcare, Security, Challenges, and Threats

Abstract Over the last four years, machine learning has undergone a significant shift due to concerns over privacy and the desire for deep learning. New methods of implementing machine learning, such as federated learning (FL), are replacing centralized systems and on-site analysis. FL is a decentralized approach that safeguards privacy by storing raw data on devices and using local machine learning to reduce communication costs. A central server collects and distributes shared models and knowledge among participants. Before delving into FL, this essay compares and contrasts various ML-based deployment strategies. This paper presents a unique classification of FL challenges and research areas, unlike previous analyses in the field. It is based on a thorough analysis of key technological roadblocks and current activities, and covers intricate subjects, contributions, and trends in the literature. The taxonomies encompass fundamental system models and designs, application domains, privacy and security, and resource management. Additionally, this paper discusses significant difficulties.