LSTM-based Anomaly Detection and Hybrid Security Architecture for Cyberattack Protection in Smart Cyber-Physical Power Systems

Abstract

The resilience and safety of the modern smart grids are threatened as there is a higher sophistication of cyberattacks on cyber-physical power systems (CPSS). The low accuracy of the detection and high-level false-opportunities are caused by the fact that conventional machine learning-based intrusion detection systems often fail to simultaneously elicit temporal relations and geographical associations with attack information. To address these limitations, this paper proposes a hybrid deep learning model, which is a mixture of Long Short-Term Memory (LSTM) network as a temporal sequence modeller and Convolutional Neural Networks (CNN) as a spatial feature extractor. The experiments of a simulated dataset of CPPS demonstrate that the proposed hybrid model is superior to the existing methods, which have an accuracy of 81 to 91, with an accuracy of 94. These findings prove that the LSTM-CNN algorithm provides a reliable and effective security system against adaptable cyberattacks in smart CPPS.