Adaptive Trust-Driven Federated Learning with Blockchain for Secure AI Healthcare Diagnostics

Healthcare supply chains and IoMT-driven clinical monitoring increasingly rely on real-time data exchange, yet remain vulnerable to adversarial attacks, counterfeit drug distribution, and privacy violations. Traditional federated learning improves data confidentiality but struggles with malicious nodes, spoofing, and unreliable participation. To address these gaps, this work proposes the Adaptive Trust-Driven Blockchain Federated Learning (ATB-FL) framework, which unifies behaviour-based trust evaluation, blockchain authentication, and incentive–penalty mechanisms into a scalable security architecture. The framework ensures tamper-resistant traceability, dynamic participant validation, and regulatory compliance while preserving data locality. Experiments conducted on the CIC-IoMT 2024 dataset demonstrate that ATB-FL improves diagnostic accuracy to 95.1%, reduces misclassification below 5%, and enhances blockchain throughput by more than 30% compared with conventional baselines. Practical evaluations, including vaccine cold-chain monitoring and remote diagnostics, further validate its applicability. These findings position ATB-FL as a trustworthy foundation for next-generation secure and privacy-preserving healthcare ecosystems.