HDRL-IDS: A Hybrid Deep Reinforcement Learning Intrusion Detection System for Enhancing the Security of Medical Applications in 5G Networks

The evolution of next-generation wireless networks, particularly the integration of Multi-access Edge Computing (MEC) in 5G, is set to revolutionize the infrastructure of secure systems. This is exemplified in the Internet of Medical Things (IoMT) field, where benefits such as remote surgeries and diagnostics become increasingly common, especially in pandemic scenarios. However, incorporating MEC services into the 5G framework significantly enlarges the network's vulnerability to traditional security breaches and introduces new, sophisticated types of attacks. These emerging threats, often undetectable by conventional methods, necessitate the development of an adaptive Intrusion Detection System (IDS) capable of identifying such complex security challenges with minimal human intervention. To tackle these issues, this paper introduces a novel Hybrid Deep Reinforcement Learning IDS (HDRL-IDS), rooted in an actor-critic methodology, designed to detect more complex security threats adaptively and autonomously with limited human oversight. Our HDRL-IDS combines the analysis of both network and host features, effectively leveraging the advantages of both Network IDS (NIDS) and Host IDS (HIDS). Empirical results indicate that our HDRL-IDS outperforms traditional NIDS and HIDS in threat detection efficacy. Furthermore, we present a novel dataset derived from an emulated 5G testbed with integrated MEC services to facilitate advanced research and development in the field, specifically for applications designed to address intricate attacks and scenarios demanding high reliability.