一种面向云计算的网络异常流量分组方法  

Network abnormal flow grouping method for cloud computing

作　　者：李春林[1] 黄月江[1] 牛长喜[1] 

机构地区：[1]中国电子科技集团公司第三十研究所,成都610041

出　　处：《计算机应用研究》2014年第12期3704-3706,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(61271171;61001084)

摘　　要：针对云计算下大量用户环境下网络异常流量分组管理问题进行研究,提出一种面向云计算的网络流量分组方法。该方法采用BIRCH算法对云计算产生的大量网络数据进行初次聚类,然后根据初次分类的统计特征进行分组融合,从而得到对网络流量的最终分类结果。实验通过一个具有突发异常流量的仿真场景,从分组效果和基于该分组的流量管理两个角度进行验证。实验结果证明,基于该方法能有效对异常用户进行分组,并且在有效阻断异常流量的同时能极大地减少对正常流量的影响。The question of how to manage the abnormal flows among the enormous user flows is issued to cloud computing.Considering that traditional grouping methods are not suitable for cloud computing,this paper proposed a method of grouping these abnormal flows into group. The main idea was clustering the flow data of each user by BIRCH algorithm at first. And then it merged these clustering results into new groups. The merging step overcame the deficiency of BIRCH’s lacking of soft clustering. It applied this method to a scenario with abrupt abnormal flows. The result shows that the method can successfully distinguishing these abnormal users by group from normal user groups. And the result also proves that drop packets randomly within groups generated by this method having better performance than dropping packet randomly or dropping randomly weighted by service type.

关 键 词：BIRCH算法 云计算 流量分组 分层聚类 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]