SHIELD-Health: Secure Healthcare IoT with Energy-efficient Ledger-based Distributed Federated Learning

Healthcare Internet of Things (HIoT) has revolutionized patient care through continuous monitoring and personalized treatment, but it introduces critical challenges in privacy protection, data security, and resource management across heterogeneous devices. Traditional centralized machine learning (ML) approaches face significant limitations due to privacy regulations and security concerns, leading to the emergence of federated learning (FL) and blockchain (BC) as complementary solutions. While FL enables collaborative model training without sharing raw data, and BC provides immutable verification and secure record management. We present SHIELD-Health, a novel framework that synergistically integrates these technologies to create a comprehensive solution for secure analytics in healthcare environments, featuring four key innovations: (1) resource-aware computation that dynamically adapts to device capabilities (2) a multi-layered privacy architecture designed for differential privacy and secure aggregation (3) Byzantine-robust aggregation ensuring model integrity under adversarial conditions, and (4) healthcare-specific optimizations including temporal attention mechanisms for physiological time-series data. Extensive evaluation demonstrates exceptional performance across multiple dimensions, maintaining high accuracy while achieving substantial communication efficiency and energy savings for resource-constrained devices. The framework also shows remarkable resilience against poisoning attacks, and robust performance under challenging non-independent and identically distributed (IID) data distributions common in healthcare scenarios. It represents a significant advancement in privacy-preserving collaborative analytics for sensitive medical applications where security, privacy, and resource constraints are paramount considerations.