Federated Learning in Real-Time Medical IoT: Optimizing Privacy and Accuracy for Chronic Disease Monitoring

Abstract

The rising occurrence of long-term illnesses requires inventive and effective healthcare solutions, and the incorporation of Internet of Things (IoT) technologies holds significant potential in revolutionizing conventional medical monitoring. This study presents an innovative method called Adaptive Federated Learning for Chronic Disease Prediction (AFL-CDP), which is specifically designed for real-time medical Internet of Things (IoT) applications. The main objective is to enhance both privacy and accuracy in the surveillance of chronic diseases. AFL-CDP utilizes federated learning, a decentralized approach to machine learning that allows for model training on multiple edge devices without the need to transfer raw data to a central server. This not only mitigates privacy concerns related to sensitive medical data but also improves the precision of predictive models by assimilating information from various data sources. The AFL-CDP adaptability enables the ongoing improvement of the predictive model using changing patient data, resulting in personalized and timely forecasts for chronic diseases. In order to improve privacy in IoT devices with limited resources, the study integrates the utilization of SPECK, an advanced technique for preserving privacy. SPECK utilizes secure aggregation and encryption mechanisms to safeguard patient data throughout the federated learning process, guaranteeing confidentiality while preserving the integrity of the model. Ensuring data security and patient privacy are of utmost importance, particularly in the field of medical IoT. The proposed methodology is assessed using a dataset that consists of real-time medical Internet of Things (IoT) data for the purpose of monitoring chronic diseases. The model's performance is evaluated using the Area Under the Curve (AUC) accuracy metric, and AFL-CDP achieves an impressive AUC accuracy of 94.37%. This showcases the efficacy of the federated learning framework in capturing the fundamental patterns in varied and decentralized healthcare data. To summarize, this study presents an innovative and strong approach for real-time medical Internet of Things (IoT) applications, highlighting the significance of privacy and precision in monitoring chronic diseases. The combination of AFL-CDP and SPECK offers a thorough method that not only satisfies the strict privacy demands of healthcare data but also achieves a high level of predictive precision, establishing the basis for enhanced patient results and personalized healthcare interventions.