An ensemble learning for anomaly identification in SCADA system

Abstract

As automation being on the surge, SCADA (Supervisory Control and Data Acquisition) substations are driven towards unmanned operations. Security of such substations is a major point of concerned in power system environment. Communication at substations can be inferred from packet level for different purposes like establishing performance pattern of peripheral devices, anomaly detection, and threat identification. The major contributions of this paper, we have implemented ensemble learning methods to identify the anomaly in SCADA traffic on an in-house developed industrial compliant test bench. Anomalies are detected by using deep packet inspection of SCADA network traffic. A mathematical model is defined for stream based anomaly detection for SCADA traffic, and the results shows the performance evaluation Decision tree and Random forest algorithm for anomaly detection.