Research On Encrypted Traffic Application Recognition Technology Based On Deep Learning

With the development of network communication technology and the increasing awareness of user privacy protection,the proportion of encrypted traffic on the network has gradually increased.At the same time,the widespread application of encryption technology in network traffic brings great challenges to network management.How to identify encrypted traffic with fine granularity in complex networks has become an urgent problem for researchers and related organizations.Therefore,the application identification technology based on encrypted traffic has become a research hotspot in academia and industry at home and abroad.Based on the characteristics of traffic data and deep learning technology,this article draws on previous experience and improves and innovates the three parts of feature extraction,feature fusion,and algorithm modeling in encrypted traffic application recognition technology,as described below.:(1)Based on the one-dimensional sequence characteristics of the original encrypted traffic data,a feature extraction method for the original encrypted traffic word sequence is proposed.On the premise of retaining the original file format of the data packet(PCAP file),the method uses all unencrypted byte information as the identification feature after excluding application layer encryption interference.(2)Aiming at the modeling stage of encrypted data packets,a one-dimensional to one-dimensional encrypted traffic recognition algorithm is proposed.This algorithm converts data packets into Packet Bytes Matrix(PBM),and uses Attention-based Convolutional Neural Networks(ATT-CNN)and Long-Short-Term Memory Network(Long and short memory network(LSTM)respectively learn its local spatial characteristics and global timing characteristics.(3)Aiming at the optimization method of packet feature extraction,a packet feature fusion method based on attention mechanism is proposed.After converting the PCAP file to PBM,the feature vector of the data packet and the weight vector are fused for the first time,and the calculation process of the attention mechanism is optimized to highlight the important feature information in the data packet.The scheme proposed in this paper can effectively identify encrypted traffic to specific applications.Experiments show that compared with the existing algorithms for encrypted traffic recognition,the algorithm proposed in this scheme has improved accuracy and efficiency.