Research On Malicious Data Attack And Defense Of Ubiquitous Power Internet Of Things

In recent years,with the continuous maturity of information technology,the power grid is gradually developing towards the direction of intelligence.The country has officially begun to comprehensively build the "three types and two networks".In this case,the cyberspace security situation has become more and more severe.This kind of malicious data attack in UPIoT is very difficult to prevent and has strong concealment,which can easily lead to malicious tampering of power system data.Therefore,it is necessary to systematically study the malicious attacks and preventive measures of the ubiquitous power Internet of Things,which not only helps to ensure the security of power system data,but also provides support and reference for the normal operation of the grid.Based on the load redistribution attack,a malicious data attack method ubiquitous in the power Internet of Things,the thesis studies and analyzes the attack principle of the load redistribution malicious data attack,and uses the improved SqueezeNet method to detect the attack,and The MisGAN method,a generative adversarial network that learns from incomplete data,is used to prevent this type of attack,The specific work is as follows:(1)The research analyzes the principle of load redistribution malicious data attack under the ubiquitous power Internet of Things,sorts out the principle and process of power system state estimation,and analyzes the difference of the objective function in the construction of attack vectors for different attack purposes.The mathematical formula shows that this malicious data attack can bypass the residual detection.(2)In order to solve the problem of increasing data scale and large amount of network parameters,this paper combines a lightweight convolutional neural network to construct a malicious data attack detection model.This model selects the SqueezeNet method and combines the gated recurrent unit for fusion and improvement,and realizes the detection work on the edge intelligent equipment of the ubiquitous power Internet of Things.It greatly improves the efficiency of network training,optimizes the network structure,and can support the extraction of spatial and temporal characteristics of power grid data.Experiments were carried out on the IEEE standard bus system to verify that the method has high detection accuracy.(3)A defense model for malicious data attacks is proposed,using MisGAN method to achieve malicious data attack defense,using malicious attack data and historical normal data to complete the training of the network generator and discriminator,and complete the training of the internal network structure and parameter settings.Based on the IEEE standard bus system,the corresponding simulation experiment analysis was carried out,and the effectiveness and performance of the model were tested and verified.