GANs结合DNN网络的电力攻击信息识别

Recognition of Power Attack Information Based on GANs and DNN Network

为提高电力终端设备攻击信息的识别准确率,利用generate adversarial networks(GANs)生成对抗神经网络和栈式编码器对电力终端设备报文进行增强和判别,再利用深度神经网络(deep neural netwok,DNN)对报文类型进行识别分类。结果表明,所提出的方法可有效识别攻击信息,且相对于LR、KNN和RF算法,识别的AUC值明显较高,具有可行性和有效性。

In order to improve the identification accuracy of attack information of power terminal equipment,generative adversarial networks(GANs) and stack encoders are used to enhance and distinguish power terminal device messages,and then deep neural network(DNN) is utilized to recognize and classify the message types. The results show that the proposed scheme can effectively recognize the attack information,and compared with LR,KNN and RF algorithms,the AUC value of the proposed scheme is significantly higher,which further proves the feasibility and effectiveness of the proposed method.