Enhancing SCADA Security: AI Based Approaches for Attack Detection

In the industrial sector, Supervisory Control and Data Acquisition (SCADA) systems are essential for managing Industrial Internet of Things (IIoT) networks. However, these systems have become increasingly exposed to cyberattacks targeting the communication layers embedded in industrial processes. Such vulnerabilities can cause severe disruptions in manufacturing and production environments. The ongoing digitalization of Industrial Control Systems (ICS) has further amplified these risks, emphasizing the need for robust security mechanisms such as Intrusion Detection Systems (IDS). This research aims to develop a high-precision AI-based IDS capable of protecting SCADA systems from evolving cyber threats. To achieve this, three categories of machine learning algorithms were evaluated: Deep Learning models (CNN, RNN, LSTM), Boosting algorithms (XGBoost, GBoost, AdaBoost), and classical methods (RF, DT, KNN). Extensive experiments were conducted using two benchmark SCADA datasets, WUSTL-IIoT-2018 and WUSTL-IIoT-2021. The results demonstrated outstanding detection performance, with all models achieving accuracy rates above 99.91%. Specifically, RF, DT, KNN, and XGBoost reached perfect accuracy (100%) on the WUSTL-IIoT-2018 dataset, while XGBoost, LSTM, and CNN achieved 99.99% accuracy on WUSTL-IIoT-2021. Additional evaluation metrics, including precision, recall, and F1-score, confirmed the robustness of the models. The findings highlight the potential of AI-driven IDS solutions to enhance the security and resilience of industrial SCADA infrastructures.