Web Based Cyber-Attack Detection for Industrial System Using Virus Spread Optimization and G-LSTM Algorithm

Modern supervisory control and data acquisition (SCADA) models include a variety of networks in addition to Internet technologies such as big data, cloud computing, Internet of Things (IoT) and web services. Real-time notifications from these technologies enable administrators to access the system from several platforms and carry out various intricate control algorithms. However, the combined interconnected systems have increased susceptibility to various cyber-attack types. In recent years, several high-profile attacks have targeted critical infrastructures such as transportation, electrical, nuclear reactors, and water distribution systems. For this reason, researchers and operators working to create high-performance detection processes for SCADA-based systems have cyber security as a primary priority. Hence, security must be considered for web-based cyber-attack detection in industrial systems. The website log files are gathered and collected in this developed architecture as a dataset. Preprocessing has been done on these datasets to normalize the data using the Z-score normalization algorithm. The KNN-based virus spread optimization algorithm is employed to determine the optimal features from the dataset. The PCA technique reduces the prediction model’s complexity while maintaining the best possible data selection. Finally, the G-LSTM algorithm trains and evaluates the feature-reduced data to forecast Web-based cyber-attacks. Evaluated performance metrics such as Accuracy, Precision, recall and Error values are 97, 97, 98, and 3%. Thus, the Web-based cyber-attack detection for industrial systems using virus spread optimization and the G-LSTM algorithm performs better than the existing model.