论文部分内容阅读
为快速将网络应用的流量进行分类,以AucklandⅡ和UNIBS两个数据集的网络流量包为研究对象,选取网络应用程序流量中最初的8个有效载荷大小作为识别特征进行研究.由于这类特征可在早期流量阶段快速提取,因此效果显著.通过将早期载荷大小可视化的方式,分析了不同网络应用的行为模式.分析结果表明,多数网络应用程序可通过早期有效载荷大小显示出它们特有的行为模式,根据早期有效载荷大小的信息可对流量进行有效识别.在此基础上,选用3种典型的机器学习分类器,即朴素的贝叶斯分类器、朴素的贝叶斯树和径向基函数神经网络进行验证分析.实验结果显示,早期有效载荷大小可作为特征对流量进行有效识别.
In order to quickly classify the traffic of network applications, the network traffic packets of two data sets of AucklandⅡand UNIBS are taken as the research object, and the first eight payload sizes in the network application traffic are selected as the identification features for research. The results were significant for rapid retrieval during the early flow phase, and behavioral patterns of different web applications were analyzed by visualizing the early load size.The analysis showed that most web applications can show their specific behavior patterns with early payload sizes , Based on the information of early payload size, the traffic can be effectively identified.Based on this, three typical machine learning classifiers are selected, that is, naive Bayes classifier, naive Bayesian tree and radial basis function Neural network to verify the analysis.The experimental results show that the early payload size can be used as a feature to effectively identify the flow.