Power IoT Attack Samples Generation and Detection Using Generative Adversarial Networks

With the rapid development of the electric power Internet of Things, the problem of a large number of electric power Internet of Things(power IoT) terminals accessing to the power grid and the problem of “blurred borders” for the power grid is becoming serious. The security problem of the electric power Internet of Things has become a hot topic of current concern. The attack detection method based on deep learning is an effective attack detection method, but the lack of power IoT attack samples leads to the unsatisfactory effect of the deep learning model, so this paper proposes a sample generation method for power Internet of Things attacks based on adversarial generation networks(GAN), then mix the different amount of generated attack samples with the original samples, through the deep neural networks(DNN) for attack detection, and finally compare the effect of the model. This method proves that increasing the attack sample of power Internet of Things greatly improves the accuracy of model detection, and the detection accuracy can reach 98%.