Privacy-Preserving Federated Deep Learning for Wearable Diabetes Prediction: A Review

Abstract

The widespread adoption of wearable healthcare devices has transformed chronic disease management by enabling continuous monitoring and real-time collection of physiological data. However, Traditional centralized deep learning methods need sensitive medical information to be transmitted to remote servers, leading to concerns regarding data privacy, communication overhead, latency, and regulatory compliance. One interesting distributed learning paradigm that makes collaborative training possible is Federated Deep Learning (FDL). while preserving the privacy of patient data. An overview of recent developments in Federated Deep Learning for chronic model diabetes prediction is provided in this paper using wearable healthcare devices. The paper examines privacy-preserving learning techniques, communication-efficient training strategies, and Edge–Fog–Cloud computing architectures that support secure and scalable healthcare analytics. It further discusses the current research landscape, identifies key challenges related to diverse data, constrained computer power, communication efficiency, scalability, and practical deployment, and outlines potential future research directions. Overall, this review offers a thorough comprehension of existing Federated Deep Learning approaches and highlights their potential to enable secure, efficient, and intelligent chronic diabetes prediction in next-generation wearable healthcare systems.