基于深度学习的网络流量分类识别研究  

Research on network traffic classification and recognition based on deep learning

作　　者：张家颖 杨文军 ZHANG Jia-ying;YANG Wen-jun(School of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China)

机构地区：[1]天津理工大学计算机科学与工程学院

出　　处：《天津理工大学学报》2019年第6期35-40,共6页Journal of Tianjin University of Technology

摘　　要：目前互联网上会存在海量的网络流量数据信息,这些海量的网络流量数据信息还未得到充分性的利用,如果有效的采取一些必要的方法或者手段,分析整个的网络流量挖掘信息对于后期的网络发展趋势,挖掘网络当中所存在的异常状态并且有采取针对性的措施,这对于后期的网络应急响应能力的增强、抵御网络不法攻击行为、快速的维护网络空间安全等方面都具有非常重大的价值及意义.本文基于网络流量识别的基本需求,分析了深度学习经典模型-CNN的基本原理,在此基础上将原始流量进行分层处理,并建立了基于注意力机制的改进的CNN算法的网络流量识别模型,最后在国际标准数据集上进行仿真分析.实验测试结果表明,该模型可以实现对各类网络流量有效识别.For now,the Internet would be a huge network traffic data information,the vast amounts of network traffic data information has not been sufficient,if effective adopt some necessary methods or means,analysis of the whole network traffic network development trend of mining information for later,mining network of the abnormal state of existence and has targeted measures,for the later network emergency response ability enhancement,against unlawful attack,rapid maintenance of network space safety has very important value and significance.Based on the basic requirements of network traffic recognition,this paper analyzes the basic principles of the classic deep learning model--CNN,and then stratifies the original traffic,and establishes the network traffic recognition model based on the improved CNN algorithm based on the attention mechanism.Finally,it conducts simulation analysis on the international standard data set.Experimental results show that the model can effectively identify all kinds of network traffic.

关 键 词：网络流量识别 注意力机制 识别算法 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]