Federated Learning Framework for Privacy-Preserving Health Monitoring via IoT Devices

The rise of interconnected wearable devices in healthcare presents unprecedented opportunities for remote health monitoring but raises significant privacy concerns regarding patient data management. This study introduces a federated learning framework tailored for privacy-preserving anomaly detection using internet of things (IoT) healthcare devices. The proposed architecture facilitates the collaborative training of anomaly detection models across distributed datasets without requiring centralized data consolidation, thereby enhancing user data confidentiality. Utilizing data gathered from a federated IoT health monitoring dataset comprising physiological signals from 1,000 virtual patients over a 60-day period, the system demonstrates promising results in identifying health anomalies. Key evaluation metrics, including accuracy, precision, recall, f1-score, AUC-ROC, and MCC, validate the framework's efficacy in maintaining high classification performance while preserving privacy. The study's primary contribution lies in offering a scalable solution that accurately detects health irregularities, potentially reducing the risk of adverse health events, particularly beneficial in resource-limited and privacy-sensitive environments.