摘要
互联网流量的多样性和不确定性,以及大量的OTT(over the top)业务开始布局HTTPS加密[1],使得传统的流量分类技术存在识别精度不高甚至失效的问题。为解决此问题,提出了基于DFI(Deep Flow Inspection)的识别方法。通过分析不同业务的流量特点,选择适用于当前业务分类的业务特征,并结合DNS进行识别,达到了良好的区分粒度和较高的识别精度。我们使用这些特征构建DFI业务识别系统,并通过大量的识别实验和系统分析对其进行评估。结果表明,业务识别系统对各类业务的识别精度均在85%以上。
The diversity and uncertainty of internet traffic,and a great quantity of OTT business began to layout HTTPS encryption,bring a lot of problems in traditional traffic classification techniques,such as low recognition accuracies or even recognition failures.To address this problem.This paper proposes a new recognition method based on DFI,and selects appropriate business characteristics for the current business classification through the analysis of the flow characteristics of different businesses.What＇ more,combinating it with DNS,our method can distinguish granularity well and achieve a high recognition accuracy.This paper builds a DFI business identification system using these characteristics,and evaluate it by a lot of identification experiments and system analysis.
出处
《工业控制计算机》
2017年第7期65-67,共3页
Industrial Control Computer
关键词
流量特征
DFI
DNS
业务识别
flow characteristics
DFI
DNS
business identification
