Blockchain-Enabled Federated Learning in Healthcare: Survey and State-of-the-Art

Advances in Internet of Medical Things technology, information and communication technologies, and machine learning have initiated the shift in healthcare towards smart healthcare. Centralization of health data to train ML models does pose privacy, ownership, and regulatory problems. Federated learning solves such problems by distributing the learning process to several devices, but it also encounters problems like encouraging participants and model aggregation correctness. Combining blockchain and FL can solve such problems through a decentralized approach that provides greater security and privacy for intelligent healthcare. This survey provides a systematic review of blockchain-based federated learning (BCFL) systems in healthcare. Key design features of BCFLs are analyzed, such as consensus protocols, crypto protocols, storage topology, and integration processes relevant to healthcare use cases. Characteristics such as convergence delay, computation overhead, accuracy loss when privacy is an issue, and ledger scalability for different implementations are compared among common implementations. The works of recent FL-based healthcare frameworks have been discussed, along with determining the challenges and research directions for healthcare use cases.