Publication

Abstract : Strong security measures are required as Cyber-Physical Systems (CPSs) have developed into extremely sophisticated and complicated systems. However, classic anomaly detection approaches are facing difficulties because to the increase of sophisticated assaults and the growing complexity of CPSs. Furthermore, a major obstacle is the increase in data volume, which necessitates domain-specific expertise for efficient analysis. To address these issues, several deep learning-based anomaly detection systems have been created. In this work, we provide a unique approach to anomaly detection by the combination of a powerful deep learning technique, Convolutional Neural Network (CNN), with a Gaussian-Mixture Model (GMM) based on the Kalman Filter and ontology (KF). Our proposed method seeks to detect and distinguish odd behavior in CPSs. The framework has two crucial processes. The first stage in data preparation is to securely prepare the original data for usage in a new format. We provide our second introduction, the GMM-KF-BiLSTM integrated deep CNN model for anomaly detection, which enables precise estimation of posterior probability for both anomalous and normal events in CPSs.