Cyber Attack Detection for Internet of Health Things through Federated Deep Learning Technique
Liyakathunisa, Zoya Riyaz Syeda, Riyaz Sohale Syed · 2024
Recently, the Internet of Health Things (IoHT) has been adopted across the globe by numerous healthcare industries due to beneficial features such as effective drug management and treatment computerization, which results in patient satisfaction and decreased cost. Most IoHT devices have not been effectively built to protect themselves against internet threats, making them vulnerable to hackers. However, these devices are prone to cyber attacks, posing threats to the healthcare system and patient safety. Furthermore, the existing artificial intelligence solution for detecting cyber attacks in healthcare applications is based on centralized learning, in which data from remote places is sent to a single server for processing and storage, which can cause significant network bottlenecks. This study proposes a scalable, decentralized, federated deep learning framework for detecting various types of malicious cyber attacks on healthcare data. A Gated Recurrent Unit (GRU) deep learning model has been successfully trained on ECU-IoHT data using a decentralized client-server-based federated learning methodology to detect different types of cyber-attacks. The results showed that the proposed model accurately detected five types of cyber attacks with an accuracy of 99.65%. Hence, the proposed federated deep learning has the potential to minimize the risks of private data leakage and build a secure, generalizable, low-cost, and scalable system for cyber attack detection for healthcare data.