Evaluation of data preprocessing techniques for anomaly detection systems in industrial control system

Abstract

The critical infrastructure can be defined as main cornerstone of modern society. Therefore, the cyber protection of critical systems like industrial control systems is vital for every modern state. However, conventional techniques are often ineffective to protect these systems. Thus, machine learning is an exceptional way to ensure cyber security in the case of critical infrastructure. The machine learning can process high dimension datasets with thousands of record in real-time. However, these datasets have to be in a proper format. The data preprocessing is a crucial stage in machine learning and can negatively influence final results. We introduce a comprehensive comparison of the main data preprocessing techniques in the relation of the network anomaly detection system. Moreover, the preprocessing of continuous datasets is considered as the subject of the research The neural network autoencoder is considered as an anomaly detection algorithm which is used to evaluate proposed solutions.