Research And Implementation Of Network Traffic Classification System Based On Deep Learning

With the continuous development of the informatization of the entire society,the scale of the Internet is also increasing.Researching network traffic classification algorithms and establishing a corresponding network traffic classification system can produce huge social value and economic benefits,and have great significance for the development of communication networks.With the continuous development of Internet technology,various new types of network applications are constantly being developed,and they all have their unique flow characteristics,which makes the characteristics of the network flow data transmitted in the communication network more complicated.Effective classification and management of network traffic have also become increasingly difficult.Through the network traffic classification technology,the traffic in the communication network is classified and managed,so that the behavior patterns of Internet users can be accurately described,the future development trend of the Internet can be accurately analyzed,and the dynamic management of network resources and the improvement of utilization rate can be achieved.The main contents of this paper are as follows:1.We propose a network traffic classification algorithm based on convolution attention.On the one hand,the traditional network traffic classification method mainly relies on the matching mode at the application layer of the network traffic,which leads to the failure of the classification of encrypted traffic.On the other hand,methods based on machine learning require domain experts to design characteristics of network traffic data,which makes it difficult to handle complex network protocols.Therefore,this paper proposes a network traffic classification based on a convolutional attention network.It first uses the attention mechanism to capture the importance of different bytes through network traffic.Then,use CNN to learn traffic features,and input the feature representation into the classifier to obtain the final classification result.It makes it possible to learn enough information from network traffic data and ensure the accuracy of classification.A large number of experiments on public network traffic data sets prove the effectiveness of the model.2.We propose a network traffic classification algorithm based on graph convolutional networks.The available information for network traffic classification consists of two parts:complex traffic internal characteristics and diverse network-side behaviors.In order to make full use of these two parts of information,this paper proposes a network traffic classification method based on graph convolutional networks.This method first uses the convolutional neural network to learn the internal characteristics of the flow,then uses the graph convolutional network to learn the interactive information on the network side,and finally inputs the learned flow embedding representation into the classifier to obtain the classification result.By fusing the internal characteristics of the traffic and the interactive information on the network side,the classification performance of the model is improved.3.We design and implement an analysis prototype system for network traffic.The task of network traffic classification requires not only the support of algorithm theory but also the network traffic analysis prototype system as a carrier for landing.This paper designs and implements a network traffic analysis prototype system based on deep learning technologies such as convolutional neural network,attention mechanism,and graph convolutional network,and system development tools such as SpringBoot and Vue.js to help communication network business.The personnel realize the analysis of network traffic and make network resource management decisions based on the analysis results.