Traffic anomaly detection method based on bidirectional autoencoder generative adversarial network

Abstract

This paper proposes a traffic anomaly detection method of bidirectional autoencoder generative adversarial network. This method first designs a twice encode autoencoder, and then fuses the twice encode autoencoder with the bidirectional generative adversarial network to solve the cycle inconsistency problem of the bidirectional generative adversarial network. Then, in order to make the bidirectional autoencoder generative adversarial network obtain more excellent detection effects, this paper improves the threshold selection method, so that it can fine-tune the threshold by setting the value of hyperparameters on different data sets, making the setting of the threshold more flexible. Finally, experiments on the UNSWNB15 dataset and KDDCUP99 dataset prove the effectiveness of the proposed method in solving the cyclic inconsistency problem of bidirectional generation adversarial networks and the superiority of the intrusion detection method proposed in this paper.