A New Dataset for Intrusion Detection in Industrial Control System: A Gas Pipeline Testbed Study
Longmin Deng, Xuemin Zhang, Qianrong Zheng, Dongdong Zhao, Junwei Zhou, Jianwen Xiang · 2023
The security of Industrial Control Systems (ICS) is a critical issue that has gained increasing attention in recent years. Machine learning based intrusion detection systems (IDS) have shown promise in detecting previously unknown attacks, but their performance depends heavily on the quality of the dataset. However, most existing datasets in the field are either outdated or do not reflect the specific characteristics of ICS. We construct a dataset of ICS network traffic collected from a gas pipeline testbed, which covers a range of normal and abnormal events specific to the domain. We detail our methodology for constructing the dataset based on the gas pipeline testbed. Moreover, the systematic and fundamental comparative experiments are performed, based on the collected dataset. Specifically, we evaluate the performance of five supervised machine learning models on it while comparing their results. Our experimental results show that the dataset can serve as a valuable resource for data-driven IDS research and that Random Forest is an effective model for detecting anomalies in ICS data.