A Federated Learning Based Privacy-Preserving Smart Healthcare System
Jiachun Li, Yan Meng, Lichuan Ma, Suguo Du, Haojin Zhu, Qingqi Pei, Xuemin Shen
Abstract
The rapid development of the smart healthcare system makes the early-stage detection of dementia disease more user-friendly and affordable. However, the main concern is the potential serious privacy leakage of the system. In this article, we take Alzheimer's disease (AD) as an example and design a convenient and privacy-preserving system named <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ADDetector</small> with the assistance of Internet of Things (IoT) devices and security mechanisms. Particularly, to achieve effective AD detection, <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ADDetector</small> only collects user's audio by IoT devices widely deployed in the smart home environment and utilizes novel topic-based linguistic features to improve the detection accuracy. For the privacy breach existing in data, feature, and model levels, <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ADDetector</small> achieves privacy-preserving by employing a unique three-layer (i.e., user, client, cloud, etc.) architecture. Moreover, <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ADDetector</small> exploits <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">federated learning (FL) based scheme</i> to ensure the user owns the integrity of raw data and secure the confidentiality of the classification model and implement <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">differential privacy (DP) mechanism</i> to enhance the privacy level of the feature. Furthermore, to secure the model aggregation process between clients and cloud in FL-based scheme, a novel <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">asynchronous privacy-preserving aggregation framework</i> is designed. We evaluate <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ADDetector</small> on 1010 AD detection trials from 99 health and AD users. The experimental results show that <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ADDetector</small> achieves high accuracy of 81.9% and low time overhead of 0.7 s when implementing all privacy-preserving mechanisms (i.e., FL, DP, and cryptography-based aggregation).