SECURE FEDERATED LEARNING FRAMEWORK FOR MEDICAL DATABASE USING BLOCKCHAIN AND CRYPTOGRAPHY

The growing application of federated learning (FL) in healthcare provides a promising solution to collaborative model training without raw patient data sharing. Yet, current FL systems are plagued by severe challenges, such as privacy threats during model aggregation, unverifiability of shared updates, and susceptibility to tampering. To solve these problems, we design an FL framework that is strong, privacy-resilient, and incorporates Elliptic Curve Cryptography (ECC) for secure model updates, Secure Multi-Party Computation (SMC) for tamper-proof aggregation, and blockchain technology for immutable audit trails. We also introduce a verifiable secret sharing (VSS) scheme to provide the correctness of aggregated models without revealing local data. Our solution uses Multi-Layer Perceptron (MLP) networks to distribute disease prediction while reducing computational overhead via optimized crypto protocols. Experimental findings illustrate that the framework provides increased confidentiality of data, integrity of the model, and scalability throughout healthcare networks. This paper fills the loophole between decentralized machine learning and regulatory compliance by facilitating transparent, secure, and efficient analysis of medicaldata.

Keywords: Federated Learning, Healthcare, Blockchain, Verifiable Secret Sharing, Elliptic Curve Cryptography, Secure Multi-Party Computation, Privacy Preservation.   

 

Federated Learning Healthcare Blockchain Verifiable Secret Sharing Elliptic Curve Cryptography Secure Multi-Party Computation Privacy Preservation.