A novel privacy protection method for mHealth systems based on differential privacy federated learning

In recent years, with the development of artificial intelligence theories and the advancement of electronic communication infrastructure, the traditional healthcare system relies on is gradually transforming into a mHealth system. However, existing mHealth systems still have some problems with user data protection. The system collects users' medical data through various mobile devices and uploads them to a centralized server in the cloud for centralized processing. If the central server is untrustworthy or suffers from a single point of attack, the risk of user information leakage will be greatly increased, and more seriously, it will lead to a crisis of trust between hospitals and users. Therefore, to solve this problem, we propose a novel mHealth system based on differential privacy federation learning (DPFL-MH). Federated learning (FL) can solve the data silo problem, which protects user data privacy and enhances the scalability of the system. Meanwhile, by using differential privacy (DP) method to encrypt the model gradient uploaded by users, it can resist malicious attacks from adversaries, thus further protecting user privacy. Finally, our experimental results show that the proposed DPFL-MH can effectively protect user privacy and perform better than existing mHealth systems.