Personalized Federated Learning for Privacy-Preserving and Scalable IoT-Driven Smart Healthcare

Abstract

The swift development of Internet of Things (IoT) in smart healthcare has facilitated real-time monitoring of patients, predictive diagnoses, and smart decision-making. Traditional centralized machine learning methods, however, present great challenges in data privacy, scalability, and flexibility to diverse patient conditions. This work suggests a Personalized Federated Learning (PFL) platform for IoT-driven smart healthcare, which overcomes the challenges of adaptive model personalization, differential privacy, and optimization of communication schemes. The PFL-HCare framework proposed here allows every IoT-enabled healthcare device to learn a personalized model while drawing on global knowledge, providing better personalization without sacrificing privacy. To improve robustness against adversarial attacks and model drift, we propose secure aggregation with dynamic model adaptation. Experimental evaluations on real-world medical data sets show that PFL-HCare improves predictive accuracy by as much as 25 % and communication efficiency by 40 % over conventional FL and centralized ML models, with privacy leakage risks reduced by 60 %. The designed framework scales across various medical applications such as chronic disease surveillance, real-time anomaly detection, and personalized medication recommendation. With its robust theoretical assurances and empirical performance, PFL-HCare has the potential to become an architectural foundation for the next generation of privacy-respecting, smart, and patient-centric healthcare solutions.

Ähnliche Arbeiten