动态的SDN网络流量模式增量学习算法  

Dynamical Incremental Learning Algorithm of Traffic Model for Software-Defined Networks

作　　者：杨喜敏[1] 胡明明[1] 唐菀[1] Yang Ximin;Hu Mingming;Tang Wan(School of Computer Science,South-Central University for Nationalities,Wuhan 430074,Chin)

机构地区：[1]中南民族大学计算机科学学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2018年第2期98-103,共6页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(61103248)

摘　　要：针对基于人工神经网络的流量统计特征学习算法在动态适应性和可扩展性等方面尚显不足.提出了一种基于增长型自组织映射(GSOM)的增量学习算法,对软件定义网络(SDN)数据平面交换机的流表统计信息进行持续学习,动态获取网络流量的GSOM神经网络模型.基于DARPA99数据集的实验结果表明:所提出的算法能够通过学习确认安全的SDN流量,获得稳定、可塑的流量模式,对异常流量也有较高的敏感度.The learning algorithm based on artificial neural network is still insufficient in dynamic adaptability and expansibility for describing the statistical features of network traffic. In this paper,an incremental learning algorithm is proposed based on growing self-organizing map（ GSOM）. It continuously learns the flow-table statistics from the switchers on the data flat,and obtains a dynamic incremental GSOM neural network traffic model for the software-defined network.Experimental results based on the DARPA99 data set show that the proposed algorithm can obtain stable and plastic traffic patterns through incremental learning the traffic which has been confirmed to be secure,and the traffic patterns also perform well in identifying abnormal traffic in software-defined networks.

关 键 词：软件定义网络 增长型自组织映射 流表统计信息 增量学习 

分 类 号：TP393.2[自动化与计算机技术—计算机应用技术]