Research On Intelligent Classification Method Of Network Traffic Based On Deep Learning

The transformation of network-centric warfare in the military field to intelligent warfare has promoted the development of intelligent networks.New technologies such as 5G and the Internet of Things in the civilian field have promoted the development of multi-heterogeneous networks,making network traffic heterogeneous and instantaneous.Management brings severe challenges,and it is urgent to solve the problem of classification of network traffic under the characteristics of heterogeneity and instantaneous emergence,so as to further improve the quality of network services.Aiming at the problems that traditional methods are difficult to accurately describe the spatiotemporal characteristics of network traffic,resulting in low traffic classification accuracy and low algorithm convergence speed,this paper proposes a software-defined network traffic classification method based on spatiotemporal attention.The spatial component consists of channels and spatial attention modules,and the temporal component consists of temporal attention modules and multi-layer bidirectional GRU stacking.Finally,an innovative reconstruction mechanism integrates spatiotemporal features.The simulation results show that the method in this paper has a fast convergence speed,the classification accuracy is improved by 1.1%～5.4%,and it has better classification performance under different traffic.Most of the existing mainstream deep learning methods only consider the internal statistical features of the traffic,ignoring the external interaction features of the traffic during communication,and the existing network traffic classification methods based on graph theory have the typical problem that the classification accuracy is not high due to the lack of map composition ability.A network traffic classification method based on multi-graph convolution is proposed.The method includes a spatiotemporal feature extraction component SCF-NET and an external interactive feature extraction component MGCN.Integrate internal and external features to achieve accurate classification of network traffic.The simulation results show that the classification accuracy is improved by 0.9%～2.8%.