【作者】 郭迟；

【导师】 王丽娜；

【作者基本信息】 武汉大学 ， 信息安全， 2010， 博士

【摘要】 在过去的40年,计算机网络尤其是Internet已经从一种研究兴趣的产物转变为一种社会的基础设施。Internet已然是推动科学技术革新和社会进步的强大引擎。但同时,社会对互联网的依赖与互联网本身的可靠需求越来越不相称。网络脆弱性的存在阻碍了互联网成为一个真正可信赖的、可靠和可预期的关键基础设施。本文的工作以复杂网络相关理论和方法为指导。复杂网络研究应用广泛,在社会、政治、经济等诸多领域具有理论意义,尤其在计算机网络领域取得了很大的成就,以此为基础形成的网络科学(Network Science)在网络计算理论、人类主题实验中的行为分析、网络科学与网络设计、网络设计与网络工程、网络设计与社会价值等研究方面具有先进性。网络脆弱性也是其中的研究热点之一。本文的工作主要围绕以网络传播动力学特征分析为主要方法的网络结构脆弱性研究展开。主要包括：(1)传统复杂网络研究偏重于对网络结构进行静态分析,确定网络的脆弱组件(结点或边)。比如很多经典研究认为网络中度或介数(Betweenness Centrality)高的结点最应该被保护和控制。那么,这些静态指标是否真的能够准确反映结点的重要性或脆弱性呢?事实上,事物的发展是普遍联系和相互转化的。网络结点的脆弱性并不一定静态而孤立地存在,其一部分由自身的静态拓扑属性所决定,而另一部分则可能隐含于其他的结点之内。为此,我们提出了一种网络结构中的脆弱性社团发现方法,同时提出了一种更准确评估网络结点在危害传播作用上的脆弱性新指标——超介数,以更为准确揭示网络结点在危害传播中的作用和地位,更准确评估网络结点的关键性和脆弱性。(2)网络脆弱性挖掘的目的是研究网络免疫策略。网络免疫策略的成效依赖于网络脆弱性挖掘的结果。由于网络危害爆发在时间和空间上具有不确定性,不存在免疫效用对所有情况都最优的免疫纯策略。找出网络的最脆弱结点加以免疫,是在资源受限条件下尽力阻碍网络危害爆发的有效途径。同时,对于危害在网络中传播而言,“推动传播”和“阻碍传播”的因素往往是同时存在和相互伴生的。因此,一个均衡的网络免疫策略不能单纯依靠静态的、单方面的网络结构分析而获得,而应该考虑网络中“推动传播(攻击)”和“阻碍传播(免疫)”二者对抗博弈。本文提出了一种二人常和非合作的网络均衡免疫对策模型。(3)一个完整的免疫资源部署流程可以分为信息收集、扫描探测、漏洞修复和自我推进四个阶段,其中搜索探测脆弱主机是免疫推进的关键环节。本文提出了一种基于扫描方式的网络免疫推进技术,能够在不知道网络结构的条件下,根据网络脆弱性分布具有自组织临界性的特点,动态调整扫描偏好,高效命中脆弱主机实施免疫修复。经过模型推导及仿真分析,该方法能够很好地抑制危害传播,提高网络的安全性。(4)网络的相继故障(Cascading failures)是指网络在遭受攻击或出现内部故障时,由于其内部结构和动力学的关联性而导致其他网络结点相继出现服务失效或故障的现象。相继故障是网络脆弱性的一个重要表现。在这一领域中,人们通常采用“负荷-容量”模型(Load-Capacity model)对网络相继故障建模。本文主要探讨了以下几个问题：①网络结点的容量-负荷关系在经济、技术条件下存在着怎样的制约关系,如何对其建模；②在有限资源条件下,如何实施网络容量分配,以最为有效地防范相继故障,使网络具有较高的鲁棒性。本文的研究结果有助于有限资源下的网络结构优化部署,抑制网络拥塞,避免网络相继故障。(5)通过对大样本数据的实证研究发现,网络行为活跃性存在着幂律涌现和社团效应,具有自组织临界特点；针对这种活跃性,本文研究了不同流量负荷对网络相继故障的影响。在网络安全的应急响应时,应该更关注那些原本不活跃的结点间流量的变化。同时,我们设计了一种具有流量自组织临界特点的低速率分布式DoS攻击的可能性,说明网络流量行为的活跃性存在能够为网络攻击所利用,也是网络脆弱性研究需要关注的。(6)网络仿真是基于复杂网络理论的Internet脆弱性主要研究手段之一。经典的网络仿真工具在网络结构分析、复杂网络动力学方面不太关注；而在传统复杂网络研究中,人们通常使用的数学、图形工具又缺乏对互联网行为的有效支持。本文工作将二者结合,实现了一个用于复杂网络脆弱性研究的大型网络软件仿真平台。如何支持用户定制的算法模型也是我们工作的出发点之一。本文实现了面向复杂的网络脆弱性分析所需要的分布并行仿真技术,给出了相关设计方法。本论文整体上采用了理论与实践相结合的研究方法。主要解决了网络脆弱性研究中的一些基础理论问题。研究成果可为我国信息安全建设做出贡献。更多还原

【Abstract】 Over the past forty years, computer networks, especially the Internet, have evolved from research curiosity to fundamental infrastructure in human society. The Internet has been a powerful engine for technological innovation and social evolution. However, societal reliance on the Internet is increasingly disproportionate to the ability of the internet to deliver high dependability and security. The network vulnerability prevents the Internet from advancing to become a truly dependable, reliable and predictable infrastructure.Our works related to complex networks theory and methodology as the guide. The complex networks, because of their remarkable theoretical significance, are widely used in the social, political, economic and many other fields. Especially in the field of computer networks, complex networks researches have made great achievements. Based on complex networks, a new interdisciplinary science named "Network Science" is emerging. Obviously, it is advanced for researches on the theory of networked computation; the behavior, computation and networks in human subject experimentation; the network design and the network engineering. Network vulnerability analysis is also a hot topic in Network Science.This thesis mainly concerns on the vulnerability in network structure, by means of analysis on dynamics characteristics of the network spread. The details are as follows:(1) The traditional research place more emphasis on static network structure to identify its vulnerable components (nodes or edges). For instance, many of them considered that the nodes with high degrees or betweenness should be paid more attention to protecting and controlling. Whether can these static characteristics really quantify network vulnerability accurately? In fact, the vulnerability of network nodes may not exist isolatedly or statically. It is associated with each other, assortatively or disassortatively. Therefore, an algorithm for vulnerability relevancy clustering is proposed to show that the vulnerability community effect is obviously existent in complex networks. On this basis, next, a new indicator called network "hyper-betweenness" is given for evaluating the vulnerability of network nodes. Network hyper-betweenness can reflect the importance of network nodes in hazard spread better.(2) Network vulnerability mining aims to develop immunization strategy. The effect of network immunization strategy relies on the result of network vulnerability mining. In order to prevent the hazard spread in a network more efficiently, we should deploy the limited security prevention resources to the most vulnerable nodes. There is not an absolutely effective strategy because the hazard in a network occurs ineluctably but we can’t predict where network hazard spreads from. In fact, "beneficial to spread" and "impeditive to spread", as a typical pair of contradictory in hazard spread, often exist at the same time. Therefore, a equilibrium network immunization strategy should be studied in an oppositional and gamble environment. A two-player, non-cooperative, constant-sum game model is designed to obtain an equilibrium network immunization strategy. (3) A complete process of immunity resource deployment can be divided into four stages: information gathering, scanning, bug fixing and self-propulsion. Where, search for vulnerable hosts is essential to network immunity. A network immunity technology on the basis of dynamic preference scan is presented. The strategy can select vulnerable hosts efficiently to fix them on the basis that the distribution of network vulnerabilities is self-organized and network structure is unreachable. The analysis of modeling and simulation shows that the network immunity method proposed in this thesis can restrain hazard spread efficiently and improve network security.(4) Cascading failures occur in computer networks (such as the Internet) in which network traffic is severely impaired or halted to or between larger sections of the network, caused by failing or disconnected hardware or software. "Load-Capacity" models are usually used for solving network traffic problems and exploring the mechanisms of cascading failures. This thesis discusses the following questions:①How to model the relationship between capacity and load of network nodes under the restriction of economic and technological conditions?②How to allocate the limited redundant resources to a network with a specific structure in order to improve the network robustness. We propose an evolutionary algorithm to search an optimized capacity allocation strategy, which could help the network achieving optimal robustness with the same resources.(5) It finds that the power-law exists in the distribution of network behaviors’activity according to our empirical study with large sample data. It is obvious that there is community effect in network communications. Based on this behavior’s activity, this thesis studies the impact of different traffic load modes on network cascading failures. Results show that the influence on the network survivability brought by the traffic change of those original inactive nodes is much greater than that brought by those active ones. Besides, we design a distributed low-rate DoS attack model by making use of genetic algorithms. It shows that the network behavior’s activity can be utilized by network attacks. It also needs to be concerned in network vulnerability researches.(6) Network simulation is one of the main means in network vulnerability research. Some classical network simulation tools, such as GTNetS, OPNET, NS-2, SSFNet, NETSim and so on, have made great achievements. But they seem to be lacking in concern on network structure and dynamics. Moreover, those widely used mathematical and graphical tools such as pajeck in traditional complex networks researches can not support Internet behaviors well. Therefore, this thesis implements an integrative simulation platform for network vulnerability research, taking advantages of both above two type tools. Besides, how to support customized algorithms and models in the platform is also our main motivation. Finally, the parallel simulation technology for complex network is implemented in our platform.This thesis conforms to the research method that from theory to practice. The contents in this thesis resolve some basic academic problems for network vulnerability researches. The conclusions and results may contribute to information security in our country.更多还原