Optimal And Self-correcting Filters For Distributed Data Compression In Sensor Networks With Cyber Attack

With the development of science and technology,networked systems have been gradually embedded in various key infrastructure.The networked system integrates computer,control and communication technology.Due to the distributed structure,the resource sharing among nodes can be realized.When the system status data or measurement data are transmitted between sensor nodes,the communication environment is often unprotected.While as the communication layer,the sensor network is vulnerable to network attacks owing to the internal interconnection among sensors.Therefore,in the field of security estimation,the estimation of networked systems with network attacks has become a hot research topic at home and abroad.At the same time,there are modeling and network-induced stochastic uncertainties in networked systems,and the statistical properties of network attack signals are usually unknown.Therefore,this paper will study the distributed optimal and self-tuning estimation problems with network attacks for multi-sensor networked stochastic uncertain systems.The main research contents are as follows:1.A distributed data compression state filter with attack detection is proposed for networked stochastic uncertain systems with multiplicative noises.Network attacks may occur during data transmission between sensor nodes.Under the unknown attack signals,an improved weighted sum of squared residuals(IWSSR)method is presented to detect network attack.Using the weighted least squares method,each sensor node compresses the measurement data of the neighbor node that are not attacked into low-dimensional measurements.Then,a distributed filter is obtained by updating its own filter with compressed measurements.Compared with the distributed filter without compression,the proposed distributed filter based on compressed data has the same accuracy and reduced computational burden.2.For networked stochastic uncertain systems with random deception attacks,a distributed data compression optimal state filter and a self-tuning state filter with deception attacks are presented.Random deception attacks may occur during data transmission between sensor nodes.Bernoulli random variables are used to describe the attack phenomena of deception attack signals in sensor networks.Under the known attack rates and noise variances of random deception attack signals,each sensor node compresses the measurement data received from its neighbor node by weighted measurement fusion algorithm based on the least square method.Based on the compressed data,a distributed optimal state filter is obtained.Under the unknown attack rates and noise variances of deception attack signals,the correlation function method is employed to identify the attack rates and noise variances of attack signals,and the attack rates and noise variances of the identified attack signals are substituted into the optimal state filter of distributed data compression,and then the distributed data compression self-tuning filter is obtained.The convergence of the algorithm is analyzed by the DVESA and DESA method.3.For networked stochastic uncertain systems with random mixed attacks,a distributed data compression optimal state filter and a self-tuning state filter with mixed attacks are presented.Random mixed attacks such as random deception attacks and Do S attacks may occur during data transmission between sensor nodes.Bernoulli random variables are used to describe the attack phenomena of mixed attack signals in the sensor network.Under the known attack rates and noise variances of the random mixed attack signals,each sensor node compresses the measurement data received from the neighbor node through the weighted measurement fusion algorithm based on the least square method.Based on the compressed data,a distributed optimal state filter is obtained.Under the unknown attack rates and noise variances of mixed attack signals,the attack rates of Do S attack are obtained by detection method,and then the Do S attack rates and noise variances of deception attack signals are identified by correlation function method.The detected attack rates and noise variances of deception attack signals are substituted into the optimal state filter of distributed data compression.A distributed data compression self-tuning filter is obtained.The convergence of the algorithm is analyzed by DVESA and DESA method.