OPTIMIZED BLOCKCHAIN-ASSISTED FEDERATED LEARNING FOR SECURE MEDICAL APPLICATIONS

Abstract

The rapid growth of digital healthcare systems has increased concerns regarding patient data privacy and secure collaborative learning. Federated Learning (FL) enables decentralized model training without sharing raw medical data, making it suitable for healthcare applications. However, FL remains vulnerable to model poisoning, free-riding, and aggregation manipulation attacks. Blockchain technology introduces decentralization, immutability, and transparent consensus mechanisms to address these issues. This paper proposes a BlockchainIntegrated Federated Learning (BFL) framework for secure medical image classification using chest X-ray datasets. The proposed system ensures secure model aggregation, tamper-proof updates, and enhanced trust among healthcare institutions. Experimental results demonstrate improved classification accuracy and significant reduction in attack success rates compared to traditional centralized and standard federated systems. The integration of blockchain enhances security robustness while maintaining acceptable computational overhead. The proposed approach is suitable for real-world distributed healthcare environments.