Security Control Of A Class Of Nonlinear Systems Under Typical Cyber-Attacks

This treatise takes a class of nonlinear systems as research target,which suffer from typical cyber-attack.We research the security control problems of denial-of-service attack(Do S)or deception attack on different parts of control loop,followers' output signals,sensor or actuator.The research methods include projection operator technology,separation function lemma,backstepping and dynamic surface control technology.The main research work includes the following three contents:1.An event-triggered security control scheme is proposed for a class of uncertain nonlinear systems in which communication link between controlled object and controller is attacked by Do S attack.When system is subject to intermittent denial of service attacks,the communication link is blocked and the controller cannot receive the system state.For this reason,a switched adaptive state observer is presented to reconstruct states.The problem of solving observer gain is transformed into multiple inequalities satisfying an infinite number of solutions through Schur complement.For this purpose,a finite number of switching algorithm is proposed.Based on solving attack,a relative threshold triggered scheme is designed which combined with encoding and decoding methods to reduce the number of signal transmission bits and solve the problem of limited network resources.Combined with designed observer and controller,the stability of the closed-loop system is proved based on Lyapunov theory and average dwell time method under condition of satisfying the inequality of the relationship between stability parameter and attack characteristic parameter.2.A distributed consensus security control scheme is proposed for multi-agent systems,where followers' output signals are subject to deception attacks and the followers are non-affine form.Firstly,non-affine followers affected by deception attacks are transformed into affine form by using mean value lemma and implicit function lemma.Then,according to the characteristics of the attacked,uncontaminated neighbor information is not available.For this reason,a compensator is designed to suppress the influence of deception attacks by using projection operator method.On the basis of dealing with attacks,a new adaptive parameter normative neural network weight is introduced to reduce the number of neural network adaptive parameters,and an adaptive distributed uniform security control scheme is designed based on the backstepping framework.Finally,we analyze and prove that the error system is input-to-state practically stability by choosing the Lyapunov function,and all signals of the closed-loop system are bounded.3.A distributed consensus security control scheme is proposed for multi-agent systems where state sensor and actuator with dead zone characteristics are both subject to deception attacks,in which followers are strict feedback form.1)In the case that followers' gain functions are known,uncertain function under the influence of sensor attack is separated by the function separation lemma,and combination term of the separation parameter and the attack upper bound is estimated by introducing adaptive parameter to deal with the sensor attack.The dead zone is transformed into two parts,one containing the controller and the other containing the interval information of the dead zone by using assumption.The attack on the actuator makes the part of the dead zone offset.Another adaptive parameter is introduced to estimate the combined item which contains dead zone offset part,upper bound of sensor attack and known gain,so as to offset the impact of this attack.The Nussbaum function is used to process the compound unknown gain composed of another part of the dead zone,the attack on the sensor and the known gain.It is proved that all signals of the closed-loop system are bounded under bounded initial conditions.2)In the case that followers' gain functions are unknown,it is similar to case 1),but in the last step of each follower controller design,Nussbaum function is used to deal with the new unknown gain which combined of the sensor attack,partial dead zone and unknown gain.In the above two cases,distributed output consensus security control schemes are designed with the knowledge of graph theory respectively,and it is proved that all signals in the closed loop system are bounded under the bounded initial conditions.