Detection of data injection attack in industrial control system using long short term memory recurrent neural network

摘要

In 2010, the outbreak of Stuxnet sounded a warning in the field of industrial control.security. As the major attack form of Stuxnet, data injection attack is characterized by high concealment and great destructiveness. This paper proposes a new method to detect data injection attack in Industrial Control Systems (ICS), in which Long Short Term Memory Recurrent Neural Network (LSTM-RNN) is a temporal sequences predictor. We then use the Euclidean detector to identify attacks in a model of a chemical plant. With simulation and evaluation in Tennessee Eastman (TE) process, we show that this method is able to detect various types of data injection attacks.