Detection And Countermeasure Of Replay Attacks In Networked Control System

Networked control system is a feedback control system which system components are connected via the shared network. As the insertion of network to the conventional control system, many convenience and network attacks are introduced at the same time, and the security has become a focus of government and scientific researchers. As one type of network attacks, the replay attack may significantly hamper the networked control system when happen any successful attack. Therefore, the replay attack of networked control system is studied in this dissertation. The main works are shown as follows:Firstly, dealing with the packet dropout in networked control systems may affect the replay attack detector. The random noises are added to the control signal artificially, and then an attack detector is used to judge the system is under attack whether or not. But when the dropout appears between sensor and Kalman filter, the detector may mistake it as attacks. In order to overcome the shortage of this method the states predicted by Kalman filter are used to compensate the lost information. The proposed method is verified by simulation in True Time toolbox which simulates the network environmentSecondly, a replay attack detection method based on the state of propagator is proposed. Although the method that injecting a random noise to control signals can detect the relay attack, it damages the networked control system performance as well. In order to deal with the problem, a state propagator is used to estimate the states to calculate the detection function. Thus the attack can be detected successfully with the control signal unchanged. In addition, the conditions that make the control system unstable by replay attack are concluded as well.Thirdly, a compensation controller is designed which based on multi-step predictive control algorithm to respond to replay attacks. When the system is suffer replay attacks, to avoid using the affected information, Kalman filter is no longer running the state correction step that using the state prediction value replace state estimate value. In order to reduce the estimation error which bring from single state predictive value, compensation controller use a plurality of past state value to estimate the value of current state. The control signal based on the partial optimal prediction algorithm to maintain the system stability. Thus, the networked control system avoids using the effected information, and the system performance can be guaranteed.Finally, the results in this dissertation are summarized and further work is also discussed.