A DoS attack detection method based on adversarial neural network

Abstract

In order to analyze the influence of deep learning model on detecting denial-of-service (DoS) attacks, this article first examines the concepts and attack strategies of DoS assaults before looking into the present detection methodologies for DoS attacks. A distributed DoS attack detection system based on deep learning is established in response to the investigation’s limitations. This system can quickly and accurately identify the traffic of distributed DoS attacks in the network that needs to be detected and then promptly send an alarm signal to the system. Then, a model called the Improved Conditional Wasserstein Generative Adversarial Network with Inverter (ICWGANInverter) is proposed in response to the characteristics of incomplete network traffic in DoS attacks. This model automatically learns the advanced abstract information of the original data and then employs the method of reconstruction error to identify the best classification label. It is then tested on the intrusion detection dataset NSL-KDD. The findings demonstrate that the mean square error of continuous feature reconstruction in the sub-datasets KDDTest+ and KDDTest-21 steadily increases as the noise factor increases. All of the receiver operating characteristic (ROC) curves are shown at the top of the diagonal, and the overall area under the ROC curve (AUC) values of the macro-average and micro-average are above 0.8, which demonstrates that the ICWGANInverter model has excellent detection performance in both single category attack detection and overall attack detection. This model has a greater detection accuracy than other models, reaching 87.79%. This demonstrates that the approach suggested in this article offers higher benefits for detecting DoS attacks.