Research And Implementation Of Network Application Encryption Traffic Classification Method Based On Deep Learning

In recent years,nearly 90%of the global network traffic has applied encryption algorithms to encrypt the transmitted data for the consideration of information security and user privacy protection.While improving network security,it also reduces the transparency of the current network environment and increases the difficulty of traffic control.Therefore,the classification and identification technology of encrypted network traffic has gradually become a research hotspot.However,most of the existing methods have problems such as low recognition accuracy and high computational complexity,which are difficult to be applied to the current network environment.Therefore,how to quickly and accurately identify the encrypted traffic data generated by different applications from the network,to optimize the network resource allocation and improve the network service quality becomes issue needing taking care of desperately in the field of network management planning.In order to classify encrypted traffic in networks,this paper proposes a network application encrypted traffic classification and recognition model based on deep learning.Based on this,this paper designed and implemented the traffic classification system.The main work of this paper includes two parts.Firstly,this paper puts forward a network application encrypted traffic classification and recognition model based on deep learning.This model uses the spatial,temporal,and statistical characteristics of network traffic to construct a three-dimensional feature dataset,then visualizes traffic data for multi-dimensional co-training.Using deep hierarchical convolution neural network model to learn the overall structure features of the original traffic and flow sequence communication features directly,to complete the classification and identification of encrypted traffic.Secondly,according to the proposed model,this paper designs and implements an encryption traffic classification system based on deep learning,which includes the business-layer of the classification model trained update module and flow classification recognition module,as well as the service-layer and data-collection-layer to provide services to the business-layer which includes traffic acquisition module,pretreatment module,features image generation module,visualization module,learning module.The system can train and save the optimal model on the labeled dataset,to realize online or offline identification of the application types of unknown encrypted traffic.Experimental results show that the model achieves 97.33%accuracy,on a dataset of about 120,000 samples covering 10 different applications.In addition,when compared with relevant literatures,this model has significant performance improvement on all indicators,without introducing additional time and space complexity,showing good detection accuracy and classification performance.At the same time,the performance testing verified that the system has good real-time,stability and scalability,so can flexibly adapt to network environment changes and application service updates.