Research on Network Traffic Classification Method Based on CNN–RNN

Abstract

With the continued development of the information society, the scale of the Internet grows ever larger, and the enormous amount of traffic data throughput in the Internet is so diverse that the difficulty of the network traffic classification problem is constantly challenged, and the speed and accuracy of the classifier are put forward to higher standards. It is critical to understand how to continuously optimize network traffic classification techniques using existing technologies for implementing network security censorship, strengthening network security management, detecting and defending against network intrusion, and so on. The traditional port-based classification method is no longer reliable in today's increasingly complex network environment, and the deep packet inspection technique is inapplicable to the classification of encrypted network traffic. The machine learning-based classification method has good classification performance and solves the problem that deep packet inspection cannot identify encrypted network traffic, which has been a hot research topic in network traffic classification in the recent years. To address the shortcomings of existing methods, this paper proposes a classification method combining recurrent neural network and convolutional neural network within the framework of machine learning.