轻量级的网络流量分类算法

Lightweight Algorithm for Network Traffic Classification

网络流量分类根据流量特征在流量数据与应用类型之间建立映射,是网络规划与运维管理和网络安全领域的基本工作之一[1]。由于网络技术的快速发展及网络流量的急剧上升,针对网络流量快速而精确的自动化分类是十分必要且刻不容缓的。通过MobileNet[2]、ResNet[3]、DenseNet[4]、GoogleNet[5]等经典卷积神经网络的研究,文中提出了一种轻量级的网络流量分类算法,利用残差网络的短连接及嵌入与激励模块的设计思想及结构优势完成网络流量分类任务。通过实验对比结果表明,该算法显著降低训练时间和模型大小,具有良好的网络流量分类效果。

Network traffic classification refers to the establishment of mapping between traffic data and application types according to traffic characteristics,which is one of the basic tasks in the field of network operation and maintenance management and network security[1]. Due to the rapid development of network technology and the rapid increase of network traffic,rapid and accurate automatic classification of network traffic is very necessary and urgent. Through research on the four classic convolutional neural networks MobileNet[2],ResNet[3],DenseNet[4],and GoogleNet[5],a lightweight network traffic classification algorithm is proposed based on the design ideas and structural advantages of residual shortcut connection and Squeeze-and-Excitation module. The experimental results show that the algorithm proposed in this paper performs well on the task of network traffic classification,and significantly reduces the training time and model size.