Anomaly Detection for Network Traffic of I&C Systems Based on Neural Network

Abstract

Anomaly detection is significant for the cybersecurity of the I&C systems at nuclear power plants. There are a large number of network packets generated in the network traffic of the I&C systems. There are many attributes of the network traffic can used for anomaly detection. The structure of the network packets is analyzed in detail with examples. Then, Features are extracted from network packets. An unsupervised neural network called autoencoder is applied for anomaly detection. Training and testing database are captured from a physical PLC system which simulates a water level control system. The result of the test results shows that the neural network can detect anomaly successfully.