基于网络层析成像的IP网络路由器级拓扑识别方法研究

发布时间：2018-03-19 01:25

  本文选题：图模式　切入点：匿名路由器　出处：《电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着网络技术的发展,网络成为了现代社会中最重要的基础设施之一,已经渗透到了人们生活的各个方面。网络路由器级拓扑表征了网络中路由器之间的连接关系,可为网络规划、优化、管理提供有用基础数据,也是构建可信、安全网络环境的前提。现有路由器级拓扑识别方法主要依靠traceroute等测量工具收集原始数据,通过对数据进行分析构建路由器级拓扑。但在实际中,网络中很多节点出于安全等因素的考虑,不会响应traceroute等测量工具发送的探测包,导致测量结果存在大量匿名理由器,故无法准确识别出路由器级拓扑。针对该问题,国际上有学者提出基于网络层析成像的拓扑估计方法,该方法通过在网络边缘节点之间发送探测包,然后利用统计学的方法推断出路由器级拓扑结构。该方法最大的优点是不需要内部节点协作,因此不受网络存在大量匿名路由器的限制。但是目前网络层析成像的方法只能识别出树状拓扑结构,而无法获得网状的路由器级拓扑结构。针对现有方法存在的缺陷,本文采用传统匿名路由器识别和网络层析成像相结合的研究思路,把网络层析成像估计所得的树状拓扑作为约束进行匿名路由器识别,进而构建较为完善的路由器级拓扑结构。本文主要贡献可以概括为以下两个方面:(1)提出基于图模式的匿名路由器聚类方法:匿名路由器聚类的目的是将在拓扑图上位置相近的匿名路由器聚集到一起,从而有利于设计合理的探测包发送方案,使发送的探测包能覆盖需要识别的匿名路由器。本文提出基于图模式的匿名路由器聚类方法,通过对测量得到原始数据的总结分析,归纳出了三种匿名路由器位置相近时,测量结果的图模式,根据起始节点、目的节点间的联系,确定匿名路由器的分布位置,对匿名路由器进行聚类。(2)提出基于网络层析成像的匿名路由器识别方法:网络层析成像利用这些已知路由器设计发包方法,并利用时间延迟的协方差构建树状拓扑。本文将树状拓扑作为约束条件,根据网络层析成像的特点提出匿名路由器识别准则,逐一分析树状拓扑中每条链路的匿名路由器分布情况。根据树状拓扑的约束和识别准则,得到每条路径的匿名路由器分布方程。通过迭代求解以上方程,就可以获得具体每条链路上匿名路由器的分布情况,进而识别出原始数据中的匿名路由器。本文使用理论数据和iPlane、CAIDA提供的真实网络测量数据对提出方法进行实验验证,实验结果表明提出的匿名路由器聚类算法可以有效地对匿名路由器进行聚类,提出的基于网络层析成像的匿名路由器识别方法能有效识别已经聚类的匿名路由器。因此,本文方法可以在存在匿名路由器的条件下,获得较为准确的路由器级拓扑结构。
[Abstract]:With the development of network technology, network has become one of the most important infrastructure in modern society, which has permeated every aspect of people's life. It can provide useful basic data for network planning, optimization and management, and is also a prerequisite for building a credible and secure network environment. Existing router-level topology identification methods mainly rely on traceroute and other measurement tools to collect raw data. The router topology is constructed by analyzing the data, but in practice, many nodes in the network do not respond to the detection packets sent by measurement tools such as traceroute because of security and other factors, resulting in a large number of anonymous reasons for the measurement results. Therefore, router-level topology can not be accurately identified. Aiming at this problem, some scholars in the world have proposed a topology estimation method based on network tomography, which sends detection packets between nodes at the edge of the network. Then the router-level topology is inferred by statistical methods. The biggest advantage of this method is that it does not require the cooperation of internal nodes. Therefore, it is not restricted by a large number of anonymous routers in the network. However, the current methods of network tomography can only identify the tree topology, but cannot obtain the network router-level topology. In this paper, using the traditional anonymous router identification and network tomography, the tree topology estimated by network tomography is used as a constraint to identify anonymous routers. The main contributions of this paper can be summarized as follows: 1) A graph-based anonymous router clustering method is proposed: the purpose of anonymous router clustering is to cluster the anonymous router in topology graph. Anonymous routers on a similar location come together, Therefore, it is helpful to design a reasonable scheme of detecting packet sending, so that the transmitted packet can cover the anonymous router that needs to be identified. In this paper, an anonymous router clustering method based on graph pattern is proposed, and the original data are summarized and analyzed through the measurement. In this paper, three kinds of graph patterns of measurement results when the locations of anonymous routers are close are summarized. The distribution of anonymous routers is determined according to the connections between the starting nodes and the destination nodes. Cluster of anonymous routers. (2) A method of identifying anonymous routers based on network tomography is proposed. Network tomography uses these known routers to design a method of sending packets. The tree topology is constructed by using the covariance of time delay. In this paper, the tree topology is taken as the constraint condition, and an anonymous router recognition criterion is proposed according to the characteristics of network tomography. The anonymous router distribution of each link in the tree topology is analyzed one by one. According to the constraints and recognition criteria of the tree topology, the anonymous router distribution equation of each path is obtained, and the above equations are solved iteratively. The distribution of anonymous routers on each link can be obtained, and then the anonymous routers in the raw data can be identified. This paper uses theoretical data and real network measurement data provided by iPlanean CAIDA to verify the proposed method. Experimental results show that the proposed anonymous router clustering algorithm can effectively cluster anonymous routers, and the proposed anonymous router recognition method based on network tomography can effectively identify anonymous routers that have been clustered. The proposed method can obtain a more accurate router-level topology in the presence of anonymous routers.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.05
，

本文编号：1632216