南京邮电大学学报（自然科学版）2026，Vol.46Issue(3)：32-40,9.DOI:10.14132/j.cnki.1673-5439.2026.03.004
一种基于综合特征的网络流量分类方法
A network traffic classification method based on comprehensive features
摘要
Abstract
With the rapid growth of Internet applications,traffic classification technology has become in-creasingly valuable in the fields of network management and security.Classification methods based on a single traffic feature can hardly meet the growing and complex requirements for traffic classification.Therefore,this paper proposes a novel network traffic classification approach based on comprehensive features,which utilizes a hybrid neural network to extract the global,local,and statistical features of traffic.The hybrid neural network combines a Transformer and a one-dimensional convolutional neural network(Transformer-1DCNN)in a mutually compensatory structure to extract both global and local fea-tures of traffic.Additionally,an autoencoder is employed to extract the statistical features of traffic.By fully exploiting the advantages of different features,this method enhances the accuracy and reliability of network traffic classification.Experimental results on the public dataset ISCX VPN-nonVPN2016 show that the proposed method achieves a classification accuracy of 98.3%,outperforming existing traffic clas-sification methods.关键词
网络流量分类/综合特征/Transformer/卷积神经网络/自编码器Key words
network traffic classification/comprehensive features/Transformer/convolutional neural networks/autoencoder分类
信息技术与安全科学引用本文复制引用
赵莎莎,冯向南,陈何,陶艺瑶,张娣,陆音,张登银..一种基于综合特征的网络流量分类方法[J].南京邮电大学学报（自然科学版）,2026,46(3):32-40,9.基金项目
国家自然科学基金(62471241)和江苏省研究生科研与实践创新项目(SJCX23_0293)资助项目 （62471241）
