Research On Analysis And Identification Technology Of Encryption Application Based On Machine Learning

Network traffic classification(NTC)is an important part of network monitoring system.It aims to identify the upper services,applications and user actions of network traffic.Results of NTC is significant for network monitoring,network security and personal information protection.With the rapid expansion of network traffic and the popularization of traffic encryption technology,the classification technology of port and payload is not applicable.Therefore,we study the analysis and identification technology of mobile encryption application,and construct the encrypted traffic dataset for domestic application,and carry out the encryption application and user action identification by machine learning and deep learning technology.Firstly,aiming at the problem of lacking application encrypted traffic dataset of domestic mobile terminals and the accuracy of application identification needs to be improved,a deep learning framework based on attention mechanism is proposed to identify domestic applications.By studying encrypted traffic identification technology,based on Convolutional Neural Network(CNN)and Long Short-Term Memory(LSTM),combined with attention mechanism,a identification model for commonly used domestic applications is proposed.Through experiments on dataset captured in real world,we compared the identification accuracy of this model with other application identification models,it is verified that the proposed model can effectively improve the application identification ability based on encrypted traffic.Secondly,aiming at the problem that fine-grained actions such as user behavior in encrypted data such as instant messaging are difficult to identify,a fine-grained action identification model of We Chat chat users based on ensemble learning method is proposed.By studying the encrypted traffic characteristics of users’ fine-grained actions in Wechat,based on the commonly used machine learning algorithms and the idea of ensemble learning,an ensemble learning model based on support vector machine(SVM)and random forest(RF)is proposed.Through experiments on the dataset captured in the real world,we compared the accuracy of user action identification between the proposed model and other machine learning algorithms,it is verified that the proposed model can effectively identify user fine-grained actions in Wechat.