FEDERATED LEARNING IN HEALTHCARE: BLOCKCHAIN INTEGRATION AND SMPC VERIFICATION FOR POISONING ATTACK MITIGATION

Abstract

Federated Learning (FL) has emerged as a transformative approach in healthcare AI, enabling collaborative model training across multiple medical institutions while preserving patient privacy. However, FL is vulnerable to poisoning attacks, where adversarial participants inject malicious data to corrupt the global model. To address this, we propose a Blockchain-integrated FL framework with Secure Multi-Party Computation (SMPC) verification, enhancing both data security and model integrity. Blockchain technology ensures tamper-proof auditability, enabling decentralized trust among healthcare participants, while SMPC-based verification mechanisms detect and mitigate poisoning attacks by validating local model updates before aggregation. The proposed system enhances privacy, robustness, and attack resilience, reducing the risk of compromised AI models in healthcare applications. Experimental evaluations demonstrate that the framework significantly improves model accuracy, security, and resistance to adversarial threats, making it a viable solution for secure and privacy-preserving AI deployment in medical research and clinical decision-making.