Research On Network Traffic Identification Technology Based On Deep Learning

With the rapid development of the internet,network traffic has shown explosive growth.Therefore,how to classify and identify different types of network traffic to ensure the security of cyberspace has always been a hot research topic in the field of network security.As the first step to guaranteeing cyberspace security,research on network traffic identification technology is of great significance for optimizing network architecture,maintaining cyberspace security,and improving user service quality.In this field,using advanced classification networks in deep learning as an assistant can greatly improve the efficiency and reliability of network security detection.Therefore,designing efficient deep learning algorithms for network traffic identification has become the main focus of current research.This study will focus on the following aspects:Mobile terminal devices have lower efficiency in recognizing and classifying encrypted and malicious traffic,making efficient and accurate identification of network traffic still a challenging problem.To address this issue,this paper proposes a onedimensional convolutional neural network(Convolutional Neural Networks for Hexadecimal Payloads,Hex CNN-1D)model based on hexadecimal payloads,which uses normalization processing and attention mechanism,and adds attention mechanism modules GAB(Global Attention Block)and CAB(Category Attention Block)for classifying and identifying network traffic.The proposed model in this paper can extract effective payload information from hexadecimal network traffic to identify most network traffic categories,including encrypted and malicious traffic data.Experimental results show that the proposed method achieves an average accuracy of 98.8% in four different experimental environments.The model proposed in this paper can significantly improve the identification accuracy of network traffic data.In this paper,the key features of network traffic packets are converted into grayscale format.In this study,we found that the compositional structure of original network traffic data frames and grayscale images is very similar.Based on this observation,we propose a network traffic recognition algorithm using a twodimensional Convolutional Neural Networks for Grayscale(GCNN-2D)model,which combines the latest research in deep learning for image processing.To validate the effectiveness of our proposed model,we conducted experiments using the public network datasets ISCX-VPN-Non VPN-2016 and USTC-TF2016.The experimental results show an average accuracy of 98.7% for regular encrypted traffic recognition and97.6% for malicious traffic recognition.This paper proposes a network resource traffic scheduling method based on Qo S and Qo E.This method utilizes a preset routing algorithm to obtain the Qo S parameter values of multiple network paths,and obtains the Qo E indicator values of each network path through a trained preset mapping relationship model.A target Qo E indicator value is selected,and information transmission is performed through the network path corresponding to this target value,achieving global control optimization of Qo S and Qo E.This method can improve the efficiency of network resource utilization,save traffic bandwidth costs within the data center,and save costs for related data centers and services.Considering the different performance indicators for computation and storage in various cloud edge data centers,artificial intelligence techniques such as deep learning are used to research and predict corresponding global Qo E optimization strategies in the event of bursts.Through adjustment of routing and business path arrangement,rapid dynamic adjustment of global software-defined network resources and optimization algorithms are achieved.