An Innovative Markovian Environment for Intrusion Detection in IoT Networks

Abstract

Recently, cyberattacks on IoT (Internet of Things) and IIoT (Industrial Internet of Things) networks have become increasingly complicated and more frequent. These attacks have posed considerable threats such as Distributed Denial of Service (DDoS) attacks, malware propagation, and unauthorized data access. These attacks threaten many applications, such as smart grids, manufacturing, and critical infrastructure systems. This paper proposes a novel approach to managing cyber-attacks in IoT and IIoT networks by developing a custom Markov Decision Process (MDP) environment and applying Deep Reinforcement Learning (DRL). The agent will be trained by the applied RL algorithm to create and control security policies autonomously. This study proposes a custom MDP environment suitable for the detection of intrusions in IoT and IIoT networks using DRL. To facilitate the training and evaluation of the DRL agent, the MDP environment is designed to mitigate realistic network conditions and various intrusion scenarios.