Research On State Estimation For Fault System Under Cyber-attack

In this paper,a non-fragile state estimator is designed for discrete nonlinear neural network systems with sensor faults and external network attacks,which ensures the accuracy and effectiveness of state estimation.Main tasks as follows:(1)At present,deception attacks are the most common cyber attack,which mainly modifies the integrity and authenticity of data in the system,the system receives a false data so that the executor performs wrong behavior and damages the system.In this case,there is a higher requirement for the accuracy of the state estimator.In the actual complex environment,the state estimator may not always maintain the optimal state due to various external factors,and the sensor failure will also affect the accuracy of the state estimator.Aiming at this phenomenon,the non-fragile state estimator is designed for the Markov jump discrete nonlinear time-delay neural network with sensor fault under deception attack.In order to further save the bandwidth and resources of the network,an event triggering method is adopted for sampling.By using the augmented matrix analysis method and the Lyapunov function,the sufficient conditions for the stability of the mean square exponential of the system are obtained,and the gain matrix of the non-fragile state estimator is obtained.(2)A Do S(Denial-of-Service)attack is an attack that makes system resources unusable.From a technical point of view,an attacker can fill a buffer of a user domain or a kernel domain,block shared network media to prevent devices from communicating or receive or change routing protocols.It is mainly modeled using a queue model,a Bernoulli model,and a Markov model.Due to the diversity of network attacks,actual attackers may also use multiple attack methods to attack networked systems.Based on this consideration,the non-fragile state estimator is designed for the discrete nonlinear time-delay neural network with sensor failures that are both subjected to Do S attacks and spoofing attacks.(3)Replay attack is a very special deception attack.It mainly shows that the fake data sent by the attack is not set by the attacker himself,but the data that the system itself has used.Replay attacks require a two-step action.First,you need to record the sensor or actuator data at a certain time,and then use the recorded data to replace the real-time sensor and actuator data when launching an attack.The biggest difference between replay attacks and other network attacks is that the attack signal is the data that the system has used.Based on its attack mechanism,the model of sensor data delay is used to simulate the occurrence of replay attack.The non-fragile state estimator is designed for the discrete nonlinear time-delay neural network system with sensor fault under replay attack.Finally,the matlab numerical simulation is used to prove that the designed non-fragile state estimator has a good estimation effect.