Research And Implementation Of Network Traffic Classification Based On Adaptive Convolutional Neural Network

As the network more and more deeply into people’s social life,the effective management and monitoring of the network become more and more important,and network traffic classification,as one of its basic technologies,has been widely concerned by the academic and industrial fields.At the same time,as deep learning technology has made great breakthroughs in image recognition and natural language processing,many scholars begin to apply it to network traffic classification and achieved many progresses.However,the current network traffic classification methods based on deep learning still face problems such as difficulty in adjusting hyperparameters of control model structure and difficulty in identifying unknown traffic,which limit the accuracy and automaticity of traffic classification.In order to solve the above problems,this paper focuses on the traffic classification of adaptive convolutional neural network,and mainly carries out the following work:(1)In order to solve the problem that manual adjustment of model hyperparameters is difficult,depends on experience and time consuming,this paper proposes a network traffic classification algorithm based on convolutional neural network structure optimization.Firstly,the algorithm preprocesses the network traffic data used for training and testing,and optimizes the network structure of convolutional neural network using particle swarm optimization algorithm during training,automatically generates a more appropriate convolutional neural network structure for network traffic classification,and verifies the classification results.This algorithm eliminates the experience-dependent and time-consuming process of manual adjustment of hyperparameters,makes the optimization of convolutional neural network structure more automatic,solves the uncertainty of manual selection,and improves the accuracy of network traffic classification.(2)To identify unknown traffic and improve the classification accuracy,this paper proposes an unknown network traffic recognition algorithm based on adaptive threshold.In the detection process,the algorithm identifies the unknown traffic by setting the threshold value,and uses particle swarm optimization algorithm to optimize the search of the threshold value,so as to reduce the influence of the unknown traffic in the network on the accuracy of the model.Compared with the method without any processing of unknown traffic,the accuracy of the algorithm is greatly improved in the data set with unknown traffic.(3)Based on the above research,this paper designs and implements a network traffic classification detection system,and verifies the proposed algorithm.The system supports the real-time capture,analysis,processing and classification of network data through the cooperation of several modules,and the classification results are displayed in real time through the interactive interface,realizing the real-time online network traffic classification.The system does not need to design features according to the specific environment,and can realize real-time monitoring of network traffic.It helps network managers have a clearer understanding of the current network status,so as to achieve more effective management of the current network.