A Gated Recurrent Unit based Intrusion Detection for SCADA Networks

Abstract

Industrial control systems are rapidly evolving and they process large volumes of critical information that flow through them. Moreover, the development and advances of various industrial communication systems and the ascent of Internet connectivity have caused a surge in new types of threats and intrusions. In this study we implement a Gated Recurrent Unit (GRU) Intrusion Detection System (IDS) destined to secure Supervisory Control and Data Acquisition (SCADA) networks. The GRU algorithm used in this research is coupled to the Information Gain (IG) approach for feature selection. The NSL-KDD dataset was employed to assess the performance of the IG-GRU-IDS. The results demonstrated that with only 20 attributes of the NSL-KDD, the IG-GRU-IDS achieved a validation accuracy of 99.52%, a test accuracy of 87.49% and a F-measure of 99.51 %. These results were superior to those obtained by Simple RNNs and LSTM based RNNs.