An Enhanced Model for Attack Detection of Industrial Cyber-Physical Systems

Summary: Aiming at the problem of data imbalance in industrial cyber-physical systems, so that the attack detection method based on deep learning has a low detection rate for rare attacks, this paper proposesdeveloped an enhanced model for attack detection in industrial cyber-physical systems..This model is derived from the original dataset.Select rare attack samples and use gradient penaltyWassersteindistance generative adversarial network to conduct rare attacksSample expansion, and mix the expanded samples with the original data set to form a new data set for training the multilayer perceptron to achieve attack detection..span>To test model performance, use ModbusTested as a real data set of communication protocols.The experimental results show that,Compared with widely used data augmentation methods, the attack detection augmentation model can significantly improve the ability of attack detection.

Keywords