ACapsule Q-Learning Based Reinforcement Model for Intrusion Detection System on Smart Grid

Smart grid is an innovative technology that offers efficiency, low carbon emissions, and high energy storage. However, this promising technology has several shortcomings, including limited security. In this network, Intrusion Detection System (IDS) is one of the likely targeted systems that has limited security and is prone to several cyber vulnerabilities. To address a such challenge, several studies have been proposed to detect, classify, and mitigate these attacks using Artificial Intelligence (AI) techniques, although the proposed techniques in the literature suffer from high misdetection and false alarm rates. Additionally, limited data availability motivated the researchers to use another type of AI method, namely reinforcement learning to detect and classify attacks. In this paper, we propose a deep reinforcement learning-based technique, namely Q learning and capsule network as a deep learning model to detect attacks for IDS on smart grid networks. The benchmark of CICDDOs 2019 is selected to evaluate the model in terms of accuracy, detection, misdetection, false alarm rates, training time, and prediction time. We also investigate the performance of the proposed model based on discount values of 0.001 and 0.9. The experiments demonstrate that the proposed model has acceptable results, and the model with the lower discount values provides better results.