Decentralized Blockchain-Empowered Federated Learning for Secure and Scalable Solutions for Electronic Health Record

Healthcare is the most innovative sector within emerging technologies. The purpose of this research is to present the decentralized framework using federated learning integrated with blockchain and enhance the disease prediction using the Electronic Health Records (EHRs) to enhance scalability, security, and privacy. The integration of blockchain in the proposed architecture encompasses healthcare institutions participating cooperatively in training a global health-based machine learning approach with safeguarding sensitive patient data, thus locking in privacy requirements such as HIPAA and GDPR. To further enhancement, we employed differential privacy by adding noise to model updates, hence preventing individual patient information from being inferred. Further homomorphic encryption is used to provide more security by allowing computations on encrypted data without revealing the underlying content. Entire results are then updated on a blockchain block to integrate the system to provide a decentralized, immutable ledger that records all model updates and transactions, ensuring that these updates are transparent, traceable, and tamper-proof. Our proposed system is ensuring scalability with the growing number of institutions to join the FL network without compromising performance. The results obtained provide significant improvements from the existing solutions. Secure aggregation techniques are employed to ensure that model updates from multiple institutions are combined effectively while maintaining privacy.