A Semi-supervisory Anomaly Detection Method for Industrial Networks Security
Guoyan Cao, Zhaowen Feng, Tongli Wang · 2020
The information securities of industrial networks are getting crucial for the reliability of smart and connected industrial infrastructures and systems. Although different anomaly detection methods have been proved to be effective for industrial networks, the implementations are too slow to be used in practice. In this paper, a semi-supervisory anomaly detection method is proposed to upgrade the implementation efficiency as well as the detection accuracy. The proposed method, in the first-fold, executes a manifold learning technique to reduce arbitrary network protocol data dimension down to two dimensions, then in the second-fold, K-NN classification is implemented to recognize abnormal network data. The results illustrate that the proposed method is superior to traditional anomaly detection methods on the aspects of both accuracy and time consumption.