MEDILIGHT PRIVACY-AWARE EDGE-BASED DIAGNOSIS ASSISTANT FOR MOBILE HEALTHCARE

Abstract

With the rapid advancement of machine learning, mobile users can now submit their symptoms anytime and anywhere for instant medical assessment. To meet the low-latency requirements of real-time diagnosis services, edge computing has become a widely adopted solution. However, traditional data-driven machine learning models rely on large volumes of sensitive medical data, which raises serious privacy concerns. Therefore, ensuring strong privacy protection is essential.To address these challenges, this project introduces LPME, a lightweight privacy-preserving medical diagnosis framework designed for edge environments. LPME restructures the XGBoost model within an edge–cloud architecture by using encrypted model parameters instead of raw user data. This transformation significantly reduces the computational overhead by shifting most cipher text operations into efficient plain text operations, making the system suitable for resource-constrained edge devices. Additionally, LPME enables secure and private on-edge diagnosis, ensuring both timely responses and confidentiality of user information. Comprehensive security analysis and experimental results verify that LPME is secure, effective, and computationally efficient.