Research On Detection Methods For Deception Attack On Cyber Physical System

Cyber physical system is a system which combines physical space with cyber space by means of computing,communication and control technology.It has a wide range of applications,such as smart grid,intelligent transportation and environmental monitoring.Because the development of cyber physical system relies on network communication and other technologies,its openness is relatively strong,followed by increasing security risks.When measurement data and control data of cyber physical system are transmitted,attackers can attack the cyber physical system by monitoring and tampering with the data,stealing the information of the system and destroying the normal operation process of the system.Once the system is destroyed,the losses sustained may even have adverse effects on the national economy and livelihood.The problem of deception attack to cyber physical systems became a widely concerned issue.In this paper,detection methods for deception attack to cyber physical system are studied.The specific chapters are arranged as follows:The first chapter describes the purpose and significance of the research,the outline of basic knowledge and the current research situation.The second chapter introduces some definitions that will be used later.Chapter 3 deals with a class of linear deception attacks and their detection and state estimation when only part of the sensors are likely to be attacked when the system has multiple sensors.This type of attack can ensure that the attacker can not be detected by the general chi-square detector,while maximizing the estimation error covariance of the system states,so that the system performance is destroyed to the greatest extent.In view of this kind of attack,we consider that under the condition of multi-sensor,some sensors guarantee absolute security against attack,while other sensors have the risk of linear attack.Using the information of security components,we reconstruct the form of innovation sequence to detect the risky data.If an attack is detected,the estimated value of the system state only uses the information transmitted by the absolutely safe sensor.If no attack is detected,the two parts of the information are fused to achieve more accurate state estimation.The fourth chapter considers the problem of how to detect a linear time invariant system when it is subjected to false data injection attacks.This kind of attack can destroy the state of the system and make the state estimation error of the system tend to infinite with the increase of time.At the same time,the residuals are kept within a certain range so that they are not detected by the detector.In this chapter,an attack detection method is designed.By multiplying sensor data by encoding matrix,the system can detect the attack in time.In addition,we have designed a method for calculating the coding matrix.The fifth chapter studies the false data injection attack and its detection method of output tracking system.For the output tracking system,the false data injection attack is designed for the sensor and for both controller and sensor simultaneously.Attacks can cause false output values to be traced to the set values,while the true output values are far from the set values.Analysis is made of what conditions the system parameters need to meet if the purpose of the attack is successfully achieved.Then the detection methods of sensor data attack and simultaneous false data injection attack on controller and sensor data are designed.Simulation results show that,according to our design method,when there is an attack,the system residual will increase quickly enough to detect an attacker's attack on the system in a very short time.Finally,the main work of this paper is summarized,and the problems to be solved are pointed out.