电力自动化调度系统虚假数据注入攻击检测

Detection of false data injection attack in electric power automation dispatching system

基于虚假数据注入攻击原理,采用加权最小二乘法通过计算状态变量残差误差,获取电力自动化调度系统状态估计量;利用卷积神经网络模型划分状态估计得到的电力自动化系统的虚假数据,通过前向传播和后向传播两阶段训练卷积神经网络,并更新网络参数;经SoftMax函数分类器输出的分类结果即为检测到的虚假数据注入攻击。实验表明,该算法在重放攻击发生后仍然能快速检测虚假数据注入攻击,可在注入攻击途中过滤较高概率的虚假数据,过滤能耗低、检测率高。

Based on the principle of false data injection attack,the weighted least squares method is used to obtain the state estimator of the power automation dispatching system by calculating the residual error of the state variable.The false data of power automation system obtained by state estimation is divided by convolutional neural network model.The convolutional neural network is trained by forward propagation and backward propagation,and the network parameters are updated.The classification results output by SoftMax function classifier are the false data injection attack.Experiment results show that the algorithm can quickly detect the false data injection attack after the replay attack,filter the false data with high probability in the process of injection attack,and has low filtering energy consumption and high detection rate.