Design of a Hybrid LSTM-DT Intrusion Detection System in SCADA Networks

Abstract

SCADA systems are critical infrastructures for managing and monitoring industrial processes, essential for controlling industrial operations. With the growing prevalence of cyber threats, detecting attacks on these systems poses a significant challenge. This study presents a hybrid model utilizing machine learning and deep learning techniques to detect cyber-attacks in SCADA networks. The proposed model integrates Long Short-Term Memory (LSTM) neural networks and Decision Tree (DT) models, trained on real industrial network traffic data. The hybrid model effectively detects intrusions with high accuracy, precision, recall, and F1-score, surpassing other approaches such as KNN and LSTM-CNN. Its superior ability to analyse network data and identify temporal patterns ensures robust performance. Additionally, the model has been validated in operational and real-time scenarios, demonstrating practical applicability. This research enhances SCADA system security and provides a framework for leveraging advanced machine learning models in industrial cybersecurity.