A Machine Learning Attack Resilient and Low-Latency Authentication Scheme for AI-Driven Patient Health Monitoring System

Ghaffar, Zahid, Kuo, Wen-Chung, Mahmood, Khalid, Tariq, Tayyaba, Bashir, Ali Kashif ORCID: https://orcid.org/0000-0001-7595-2522 and Omar, Marwan (2024) A Machine Learning Attack Resilient and Low-Latency Authentication Scheme for AI-Driven Patient Health Monitoring System. IEEE Communications Standards Magazine, 8 (3). pp. 36-42. ISSN 2471-2825

Preview

Accepted Version Available under License In Copyright. Download (13MB) | Preview

Abstract

The Internet of Medical Things (IoMT) and Artificial Intelligence (AI) models have transformed healthcare by enabling wireless communication for Remote Patient Health Monitoring (RPHM) services. Wireless technologies such as Wi-Fi and 6G support reliable and low-latency communication between AI models and IoMT devices. IoMT devices allow individuals to monitor their health remotely, reducing the need for hospital visits. Integrating IoMT with AI and 6G enables automated diagnostics and personalized care with reduced data transmission among involved entities. It also helps data-intensive applications achieve higher performance levels regarding throughput, reliability, low latency, and energy-efficient communication for AI-driven RPHM system. However, exchanging sensitive information over public channels makes IoMT vulnerable to potential security attacks. Designing effective and secure mutual authentication and key agreement scheme for RPHM has been challenging due to privacy and security concerns. Moreover, there is also a demand for reliable and low-latency communication for AI-driven RPHM systems. Many existing authentication schemes have limitations, including susceptibility to machine learning attacks and high latency rates. To overcome these issues, we present a machine-learning attack-resilient and low-latency authentication scheme for AI-driven RPHM. The proposed scheme utilizes a three-factor approach based on elliptic curve cryptography (ECC). It employs a one-time physical unclonable function (OPUF) to resist machine learning attacks on medical sensing devices. The scheme's security is evaluated through informal and formal analysis, demonstrating its security strength and persistence. Additionally, the scheme's performance is assessed using various metrics, confirming its superiority over related schemes and achieving a low latency rate.