An ICS Traffic Classification Based on Industrial Control Protocol Keyword Feature Extraction Algorithm

• Published in: Applied sciences Vol. 12; no. 21; p. 11193
• Main Authors: Yu, Changhong, Zhang, Ze, Gao, Ming
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.11.2022
• Subjects: Accuracy; Algorithms; Classification; Control systems; Datasets; decision tree; Decision trees; Deep learning; feature extraction; industrial control system; Industrial electronics; Internet of Things; Keywords; Latency; Machine learning; Methods; Neural networks; Periodicity; protocol traffic classification; Traffic; Traffic control
• ISSN: 2076-3417; 2076-3417
• DOI: 10.3390/app122111193

Industrial control protocol feature extraction is an important way to improve the accuracy and speed of industrial control protocol traffic classification. This paper firstly proposes a keyword feature extraction method for industrial control protocol, and then designs and implements an industrial control system (ICS) traffic classification based on this method. The proposed method utilizes the characteristics of the relatively fixed format of the industrial control protocol and the periodicity of the protocol traffic in ICS. The keyword features of the industrial control protocol can be accurately extracted after data preprocessing, data segmentation, redundant data filtering, and feature byte mining. A feature dataset is then formed. The designed ICS traffic classifier adopts decision tree and is trained with the feature dataset. Experiments are carried out on the open-source dataset. The results show that the proposed method achieves 99.99% classification accuracy, and the classification precision and classification recall rate reach 99.98% and 99.93%, respectively. The training time and predicting time of classifier are 0.34 s and 0.264 s, respectively, which meets the requirements of high precision and low latency of industrial control system.