Research On Encrypted Traffic Classification Based On Deep Learning

With the development of the Internet and the outbreak of COVID-19,various online teaching,instant messaging and entertainment applications have developed rapidly,coupled with the gradual popularity of traffic encryption technology,increasing the complexity of encrypted traffic in the network,which constantly poses new challenges for monitoring the behavioral security of applications in the current Internet environment.At the same time,traffic encryption technology facilitates many malware to hide its attack behavior,which poses a serious threat to the information security space.As one of the important technologies for monitoring network traffic and collecting network traffic information,encrypted traffic classification plays an important role in the protection of cyberspace.In the face of complex encrypted traffic data in the current network environment,most of the traditional traffic classification methods have lost its usefulness,while the deep learning method has gradually become the mainstream traffic classification method due to its advantages of strong learning ability and great classification effect.In this paper,we explore the classification of encrypted traffic data using different methods by analyzing the information of encrypted traffic and studying the deep learning models.The main research works and innovations in this paper are as follows:Aiming at the problem that the current traffic preprocessing methods cannot accurately extract encrypted traffic information,the format of the encrypted traffic data and the information contained in each part are analyzed in depth,and a new data preprocessing method is proposed to improve the classification effect.By removing a large amount of irrelevant information to accurately extract more valuable information for classification,this preprocessing method can improve the classification effect and save model training time at the same time.In view of the poor classification effect of current encrypted traffic classification methods,an encrypted traffic classification method based on Res Ne Xt is proposed,which uses a model based on the traditional onedimensional convolutional neural network and combines the ideas of group convolution,feature fusion and the hopping idea of residual network,which solves the problems of gradient disappearance and network degradation while also improves the performance of the network model,so as to realize the accurate classification of encrypted traffic.The results of several experiments show that this encrypted traffic classification method can achieve excellent results.Most of the current encrypted traffic classification methods do not make sufficient use of important feature to allow the network to pay attention to more important features adaptively.Therefore,a channelspatial domain attention module is designed and introduced into the Res Ne Xt Module to make it a Res Ne St-like structure with corresponding simplifications and improvements,thus proposing a model more suitable for the encrypted traffic classification task.The model combines the ideas of feature fusion and hopping and implements the idea of allowing the network to pay attention to important information adaptively,while also taking advantage of the one-dimensional convolutional neural network in processing one-dimensional data streams.The experimental results show that the classification experiment using this method can effectively improve the classification accuracy and F1 score,and the classification results of different types of traffic are more balanced.