Privacy-preserving healthcare monitoring for IoT devices under edge computing

Abstract

With the rapid development of the Internet of Things (IoT) technology, e-healthcare has received extensive attention because it is able to provide real-time health status feedback for users. Due to the exponential growth of health data and the expensive cost of training the healthcare monitoring model, outsourcing the healthcare monitoring service to the edge servers becomes a popular way in e-healthcare. How to accomplish healthcare monitoring without leaking users' privacy to the edge servers is the key element of e-healthcare. In this paper, we propose a privacy-preserving healthcare monitoring framework based on long short-term memory (LSTM) neural network for IoT devices, which is performed on edge servers. In our framework, to reduce the computational burden, we outsource the complicated computations to the edge servers. To protect the privacy of health data and healthcare monitoring model, we design a series of secure interactive protocols based on replicated secret sharing and use three edge servers to collaboratively implement secure inference and model training for LSTM. We demonstrate the security of the proposed protocols and give the performance evaluation. Numerical experiments show that the proposed protocols are able to produce results with fine accuracy while requiring low computation and communication overheads.