Blockchain-based Federated Learning with SMPC Model Verification Against Poisoning Attack for Healthcare Systems

Abstract: Healthcare systems generate large amounts of sensitive patient information through hospitals, labs, wearable devices, diagnostic centres, and electronic medical records. Traditional centralized machine learning approaches require sharing raw healthcare data to train intelligent models. This raises significant concerns about privacy, security, and following regulations. Federated Learning (FL) addresses these issues by allowing multiple healthcare institutions to work together to train machine learning models without transferring raw patient information. However, federated learning environments are still at risk of poisoning attacks, where malicious actors submit altered model updates that harm model performance and reliability. This research presents a secure framework called Blockchain-based Federated Learning with SMPC Model Verification Against Poisoning Attack for Healthcare Systems. The framework combines Federated Learning, Secure Multi-Party Computation (SMPC)-based verification, and Blockchain technology to create a reliable collaborative platform for healthcare intelligence. Healthcare institutions train models locally and share only model parameters instead of patient records. Local model updates undergo SMPC-based verification to flag suspicious contributions before aggregation. Verified updates are combined using federated learning techniques, while malicious updates are discarded. Blockchain technology maintains unchangeable logs to enhance transparency, traceability, and accountability. The proposed system is implemented as a Flask-based web application supported by an SQLite database. This database manages healthcare participants, federated rounds, verification records, global model information, and blockchain event history. The framework strengthens privacy preservation, boosts security against poisoning attacks, and builds trust in distributed healthcare artificial intelligence systems. Experimental implementation shows that combining federated learning with blockchain and verification methods offers a dependable and scalable solution for secure healthcare collaboration.

Keywords: Blockchain Technology, Federated Learning, Healthcare Artificial Intelligence, Secure Multi-Party Computation (SMPC), Poisoning Attack Detection, Privacy Preservation, Distributed Machine Learning, Healthcare Security, Model Verification, Blockchain Auditability.