Research On Unknown Traffic Identification Based On Self Encoder And Attention Mechanism

In today’s network era,the emergence of a large number of new devices and the development of the Internet of Things have also led to a surge in traffic data,and the unknown traffic has also increased.In order to solve the problem of stable transmission of traffic data in the network and provide excellent network services,we need not only accurate network traffic classification,but also to identify the unknown network traffic in the network,Avoid the impact of unknown traffic on the normal transmission of network data.In network traffic classification,the classification method of known class data is relatively mature.When there is unknown traffic,the current model can only identify the unknown traffic based on the learned prior knowledge of the known classes.How to better learn the characteristics of the known classes of data to lay the foundation for the identification of unknown traffic has become an urgent problem to be solved in the identification of unknown traffic.In order to realize the accurate identification of unknown traffic and reduce the impact of misjudgment on traffic classification,this paper takes the mature convolutional neural network in network traffic classification as the basis and improves it,and introduces the method of identifying unknown classes in the field of open set recognition.The main contents of this paper are as follows:1.Unknown traffic identification method based on convolutional self encoder and distance loss.In this paper,based on convolutional neural network,the network is slightly improved,combined with the self encoder mechanism,so that the network can effectively learn the characteristics of known data.According to the effective features extracted from the model,the method in the field of open set recognition is introduced,and the method based on distance loss is used to promote the compact data of the same kind in the data and the separation of different types of data.Experiments show that the model can solve the problem of unknown traffic identification.2.Unknown traffic identification method based on attention mechanism and distribution function.This paper mainly extracts the effective features of data from two aspects of space and time series to lay a foundation for model learning,so that the model can accurately classify traffic.In addition to setting a threshold for the identification of unknown traffic,the model directly scores the data according to the distribution function of the data,so that the model can directly output the unknown probability of the data.Experiments show that the model can identify unknown protocol traffic efficiently and accurately.