Research On Key Technologies Of Mobile Application Recognition

For security reasons such as protecting users’ personal privacy from disclosure,today’s mobile applications mostly use TLS / SSL and other technologies for encrypted transmission,which also makes it more difficult to accurately identify mobile applications.Therefore,for the identification of mobile applications in the encrypted traffic environment has become a research hotspot at home and abroad,this paper uses the relevant methods of deep learning to propose corresponding solutions to the problems of insufficient identification accuracy and sample size,The details are as follows:(1)This paper proposes a cnn-gru mobile application recognition algorithm based on attention mechanism.In view of the spatial and temporal correlation of mobile application traffic,the algorithm uses convolutional neural network to capture the spatial correlation of traffic,uses cyclic neural network to capture the temporal correlation of data flow,and introduces attention module to further improve the prediction accuracy,At the same time,the convolution neural network structure is improved,the small convolution kernel and high-performance activation function are adopted,and the more lightweight Gru module is used to replace the LSTM module for improvement,so that the model can ensure the accuracy and reduce the amount of computation at the same time.(2)In this paper,a mobile application recognition model based on transfer learning is proposed to solve the problems of fast update and iteration speed of mobile applications,complex and cumbersome traffic collection and annotation and so on.This method first constructs a mobile application recognition model based on residual network,and trains it through some data,and then freezes part of the structure and weight of the trained recognition model,so as to ensure the accuracy of application recognition and solve the problem of small sample recognition on the basis of reducing training parameters and shortening training time.The improvement and innovation proposed in this paper can better solve the accuracy of mobile application recognition and the recognition problem under the condition of small samples,and has good generalization ability.The last chapter of the paper summarizes and prospects the shortcomings of the model proposed in this paper and the possible improvement methods in the future.