基于卷积注意力门控循环网络的加密流量分类方法  

An Encrypted Traffic Classification Method Based on Convolutional Attention Gated Recurrent Networks

作　　者：张彦晖 吕娜 刘鹏飞 陈卓 ZHANG Yanhui;LYU Na;LIU Pengfei;CHEN Zhuo(Information and Navigation College,Air Force Engineering University,Xi’an,Shaanxi,710077,China;The 94619 Unit,Lu’an,Anhui 237000,China)

机构地区：[1]空军工程大学信息与导航学院,陕西西安710077 [2]中国人民解放军94619部队,安徽六安237000

出　　处：《信号处理》2021年第7期1180-1188,共9页Journal of Signal Processing

基　　金：国家自然科学基金(61703427)。

摘　　要：流量加密技术给流量分类带来了新的挑战,为实现加密流量的快速准确分类,提出了一种基于卷积注意力门控循环网络的加密流量分类方法。将卷积神经网络和门控循环单元相结合,针对流量数据的特点,修改卷积神经网络的池化层以提取单个数据包特征,通过注意力机制寻找单个数据包的关键特征并赋予高权重;然后采用门控循环单元提取流层面数据包间的时间序列特征,从包层面和流层面全面反映流量的整体和局部特征。实验证明该方法相对于现有方法,提高了分类准确率、实时性和训练效率。The emergence of traffic encryption technology brought new challenges to traffic classification.In order to classify the encrypted traffic quickly and accurately,an encrypted traffic classification method based on convolutional attention gated recurrent network was proposed.This model combines a convolutional neural network and a gated recurrent unit.According to the characteristics of the traffic data,the pooling layer of convolutional neural network was modified to extract the characteristics of single data packet.The key features of a single packet were found by attention mechanism and given high weight.Then,the time series characteristics between packets at the flow level were extracted by gated recurrent unit,which reflected the overall and local features of traffic from packet level and flow level.Experimental results show that this method improves the classification accuracy,real-time performance and training efficiency compared with the traditional methods.

关 键 词：加密流量分类 卷积神经网络 注意力机制 门控循环单元 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]