Al-Driven Federated Learning for Securing the Internet of Medical Things: A Critical Review of Architectures, Threats, and Blockchain-Enabled Trust

The Internet of Medical Things (IoMT) has become central to modern healthcare by enabling continuous patient monitoring and real-time clinical decision-making, yet the rapid proliferation of connected medical devices has simultaneously expanded the attack surface and introduced serious risks to patient safety. While artificial intelligence offers strong capabilities for detecting and mitigating cyber threats, conventional centralized AI approaches often require the collection and processing of sensitive patient data, thereby recreating the very privacy risks they aim to solve. Federated learning addresses this limitation by allowing multiple devices or institutions to collaboratively train a shared model without exchanging raw data; however, it remains vulnerable to malicious or poisoned model updates from untrusted participants and still faces unresolved challenges of trust and accountability. Blockchain has been proposed as a complementary mechanism to establish decentralized trust and enhance the auditability of federated environments, although evidence of its scalability and practical feasibility in real-world healthcare settings remains limited. This paper critically examines the current landscape of security solutions at the intersection of AI, federated learning, blockchain, and IoMT. Drawing on a curated body of recent research, it maps the IOMT threat landscape across network, model, and privacy dimensions; compares AI-driven, federated, and blockchain-enabled approaches; and synthesizes findings across key technical criteria including threat coverage, detection capability, privacy protection, trust management, communication efficiency, scalability, and practical deployment feasibility. The analysis reveals three persistent gaps: inconsistent reporting practices, fragmented and heterogeneous datasets, and an unresolved tension between claimed blockchain efficiency and the limited empirical support for its scalability in real clinical environments. Notably, no existing study jointly evaluates AI, FL, and blockchain under standardized conditions that reflect a realistic hospital IOMT deployment. Based on these findings, the paper proposes a research agenda focused on real-world scalability validation, regulatory-aligned privacy standardization, ۶G-integrated architectures, explainability-by-default, and quantum-resilient federated learning, with the goal of advancing IOMT security solutions that are technically robust, trustworthy, scalable, and patient-centric.

کلیدواژه ها:

Internet of Medical Things (IoMT) ، federated learning ، artificial intelligence ، blockchain ، intrusion detection ، privacy-preserving machine learning