基于机器学习的IP流量分类研究  

Study on Internet Traffic Classification Using Machine Learning

作　　者：刘琼[1] 刘珍[1] 黄敏[1] 

机构地区：[1]华南理工大学软件学院,广州510006

出　　处：《计算机科学》2010年第12期35-40,66,共7页Computer Science

基　　金：华南理工大学985科研启动经费(D606001II)资助

摘　　要：IP流量分类是Internet研究和流量工程的重要基础,近年来网络应用类别和Internet流数量在快速增长。流量分类技术不断面临新的挑战。对基于机器学习的IP流量分类方法进行了系统性研究。给出了这类流量分类方法的数学描述;通过深入研究有监督和无监督机器学习方法在流量分类中的应用,从数据预处理、模型构建和模型评估3个方面评述这类技术的研究现状,并指出存在的问题;总结得出现阶段基于机器学习的IP流量分类技术存在数据偏斜、标识瓶颈、属性变化和实时分类等4个方面的共性问题;最后展望了流量分类技术的未来发展方向并介绍了作者正在进行的工作。Internet traffic classification is one of the key foundations for research works and traffic engineering in Internet.The categories of Internet applications and the number of Internet flows increase fast last years.The technique challenges are coped with development of traffic classification all the time.A research was carried out systematically in Internet traffic classification using machine learning in the paper.A mathematical description was given for the technology of traffic classification first.After surveying the technology of traffic classification based on supervised learning and unsupervised learning,we commented on the data preprocessing,classification model and performance evaluation etc.three aspects,and indicated the shortages now in these studies.Then,four common issues were summarized,which are data skew,label bottleneck,attribute changing and classify timely respectively.At the end,the prospect and our going work in this area were pointed out.

关 键 词：流量分类 机器学习 网络流 网络测量 

分 类 号：TP393.06[自动化与计算机技术—计算机应用技术]