Answers are grounded in this paper and cite their sources.
The rapid development of healthcare Internet of Things (IoT) systems has enabled large-scale medical data collection and intelligent healthcare services. However, medical data are highly sensitive and are typically distributed across multiple institutions, leading to severe data silos and privacy concerns. Federated learning (FL) provides a promising solution by enabling collaborative model training without sharing raw data. Despite its advantages, conventional FL lacks trust evaluation and remains vulnerable to malicious model updates. Blockchain technology offers decentralized trust management and tamper-evident records, but existing blockchain-based solutions mainly focus on data sharing rather than collaborative learning. To address these challenges, this paper proposes BTFL-Med, a blockchain-driven trust-aware federated learning framework for secure medical data sharing. The proposed method introduces a storage-aware trust modeling mechanism that jointly considers storage contribution and model quality to support reliable client selection. A blockchain-based verification scheme is designed to check model-update integrity through hash anchoring and smart contract validation. Furthermore, a trust-weighted aggregation strategy is developed to improve robustness against malicious updates. To mitigate potential information leakage during local training, Gaussian noise is incorporated into local gradients. An on-chain/off-chain collaborative storage strategy is also adopted to reduce system overhead. Theoretical analysis and experimental evaluation investigate the reliability, robustness, and efficiency of BTFL-Med under the evaluated settings. The Gaussian-noise mechanism is intended to mitigate potential information leakage.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: