An Effective and Secure Privacy Diagnosis Made Online System for the E-Healthcare Ecosystem Based on Federated Learning

The electronic healthcare (e-healthcare) approach has made it easier for consumers to access medical care. User privacy, data security and the effectiveness of online treatment, however, have also sparked broad public concern. In this paper, we present a federated learning mechanism (hereafter FLM)-based online treatment approach for the e-healthcare system that is effective and privacy-preserving. The security of data used for training can be properly safeguarded by communicating estimated locally parameters of the model rather than actual data. The support vector machine (hereafter SVM) algorithm with a homomorphic cryptosystem are then used to effectively classify physiological data from patients without disclosing their identities. In addition, we develop a fresh method for recovering the SVM model’s decision function that effectively stops the parameters of the model from leaking. Experimentally-based numerical results demonstrate the proposed scheme's great efficiency. As a result, our scheme has real-world e-healthcare ecosystem applications.