透過您的圖書館登入
IP:3.149.247.132
  • 期刊

Identification and Detection of Malicious Traffic in Communication Networks with a Deep Learning Algorithm

摘要


The development of the Internet has brought convenience to users in the process of penetrating daily life; however, corresponding requirements are also put forward for network security, especially for detecting malicious traffic. This paper introduced the traffic feature extraction method and the convolutional neural network (CNN) algorithm among Internet malicious traffic detection algorithms. The particle swarm optimization (PSO) algorithm was adopted to adjust the CNN parameters in the training process in order to avoid falling into the locally optimal solution. We compared simulation experiments on the improved CNN algorithm to compare its performance under three activation functions, relu, sigmoid, and tahn. We also compared the performance of the improved CNN algorithm with support vector machine (SVM) and back-propagation neural network (BPNN) algorithms. It was found that the improved CNN algorithm had the best recognition performance when tahn was used as the activation function and also had better recognition performance than SVM and BPNN algorithms; the improved CNN algorithm could maintain a stable malicious traffic filtering effect when facing increasing traffic.

延伸閱讀