基于机器学习的网络流量分类算法

A network traffic classification algorithm based on machine learning

互联网应用的蓬勃发展产生了种类多样的网络流量。在网络技术不断进化的过程中,新型流量和流量加密技术的出现,使基于端口和基于有效载荷的传统网络流量分类算法的应用受到限制。为了实现对新型网络流量的自动分类,提出了一种基于机器学习的网络流量分类算法。通过选择特征属性和构建决策树模型,能够实现对流量级别的网络数据进行自动分类。使用网络流量分类领域的公开数据集进行训练和测试,并将测试结果与开源的机器学习平台Weka运行结果相比较,实验结果表明:所构建模型性能优良,在流量分类准确度与Weka平台相近甚至更优的前提下,大幅降低了建模时间,提高了网络数据分类的效率。

The boom of the Internet has spawned a variety of network traffic.During the continuous evolution of Internet,the emergence of new traffic and encryption technologies have limited the development of traditional network traffic classification algorithms,such as the methods based on port and payload.In order to classify emerging network traffic automatically,a network traffic classification algorithm based on machine learning is proposed.By means of feature selection and decision tree algorithm,a classification model is built to deal with the flow-level data collected from the Internet.An open data set is employed as the training set and the testing set,and the experimental results are compared with Weka,an open-source machine learning platform.It is indicated that the accuracy of the classification model is similar to or even better than the result of Weka,while the modeling time is far less than Weka and the efficiency of network data classification is improved.