Research On Refined Classification Technology For Encrypted Traffic

As people attach importance to information security,the scale of encrypted traffic has soared.In order to provide better service quality while ensuring information security,people need to finely identify network traffic.For traditional plaintext network traffic,there is a mapping relationship between its port and protocol,and the transmission content is not encrypted.Therefore,port-based and load-based network traffic identification methods are very effective;for SSL network traffic,the transmitted data is encrypted In the latter ciphertext,data of different application types may also be transmitted using the same port.For example,different application data based on the HTTPS protocol use port 443.Therefore,the above two methods cannot achieve fine division of this type of network application.With the continuous popularization of the SSL encryption protocol,more and more network video platforms are now choosing to use the SSL protocol for video data transmission.In order to improve the quality of video services,operators need to conduct Qo E evaluation of user experience,and encryption technology The application poses a higher challenge for operators to analyze and evaluate video traffic.In response to the above problems,this paper focuses on the identification of encrypted traffic based on machine learning.Based on the existing research,through the selection of traffic characteristics and the improvement of the identification model,a fine-grained and more accurate encrypted traffic is proposed.Recognition model,and also focused on the recognition of encrypted video traffic clarity.First of all,for the problem of refined identification of encrypted traffic,this paper conducts an in-depth study on the existing traffic identification methods.Aiming at the lack of precision and the lack of fine-grained identification in the existing encrypted traffic identification methods,through the analysis of the characteristics of different types of traffic As well as the comparison of existing machine learning models,a refined identification model of encrypted traffic based on random forest algorithm is proposed.This paper introduces the methods of data normalization and data standardization in the non-dimensionalization of feature engineering,random forest parameter tuning,and feature selection based on the information entropy of a single decision tree,so that the model in this paper has better recognition accuracy Degree and fine-grained.Secondly,in response to the existing Qo E evaluation of encrypted video traffic,this paper has conducted a lot of research on the existing Qo E evaluation methods of video traffic,and found that the current research on the Qo E evaluation of encrypted video traffic is mainly concentrated on the You Tube video platform,for other video platforms There is less research on encrypted video traffic.The streaming media transmission protocol used by Youtube is slightly different from other video platforms such as Tencent Video.For these online video platforms,this paper proposes a machine learning-based definition recognition method to make up for the shortcomings of the existing encrypted video Qo E evaluation methods.