A Privacy-Enhanced Federated Learning Framework for Secure Healthcare in the Internet of Medical Things (IoMT)

Abstract

Emerging as a potential paradigm to enhance healthcare delivery by means of linked medical equipment and sensors is the Internet of Medical Things (IoMT). But when aggregating and evaluating data from scattered IoMT devices, the delicate nature of medical data begs serious privacy issues. Federated learning (FL) presents a possible answer by allowing group model training without raw data sharing. This work presents a fresh federated learning architecture especially intended for IoMT environments. The main contributions are: 1) a hierarchical FL architecture customized for heterogeneous IoMT devices and edge-cloud infrastructure; 2) privacy-preserving techniques including differential privacy and secure aggregation to protect patient data; 3) Communication-efficient protocols to handle unreliable IoMT networks; 4) Techniques to address statistical heterogeneity and non-IID data in medical datasets; and 5) Incentive mechanisms to encourage participation of IoMT device. Extensive investigations on real-world medical datasets show the efficacy of the proposed framework in terms of model accuracy, communication efficiency, and privacy preservation. Our method reduces communication by 80% and offers strong privacy assurances while nonetheless attaining 95% of the accuracy of centralized learning. This paper offers a useful federated learning method to provide effective distributed machine learning with respect for privacy for next-generation IoMT systems.