Abstract: In order to solve the data integrity attack of networked control systems, a bias attack and its detection method based on the networked inverted pendulum platform sensors are designed in this paper. First, the Ettercap tool is utilized to realize network intrusion and inject false data. Next, combined with the support vector machine (SVM) method, the LibSVM classifier is used to train the four kinds of state information in the inverted pendulum system to obtain the model and classify the data. After that, the SVM method is compared with K-nearest neighbor and decision tree methods in the self-built system. Finally, the method proposed is validated on the platform. The simulation and experimental results show that the designed attack method can change the stability of the system. Compared with the commonly used machine learning method, the SVM has more advantages in the binary classification of bias attack detection and can effectively distinguish the false data in the transmission data.

Key words: networked inverted pendulum platform, bias attack, Ettercap, support vector machine(SVM), LibSVM