【作者】 张永铮；

【导师】 方滨兴；

【作者基本信息】 哈尔滨工业大学 ， 计算机系统结构， 2006， 博士

【摘要】 在Internet规模不断增长的同时,与Internet有关的安全事件也愈来愈多,安全问题日益突出。理论和实践分析表明,诸如计算机病毒、恶意代码、网络入侵等攻击行为之所以能够对计算机系统产生巨大的威胁,其主要原因在于计算机及软件系统在设计、开发、维护过程中存在安全弱点。因此,深入分析和研究安全弱点对保障计算机系统与网络安全具有重要意义。弱点描述和评估技术是弱点研究领域的两个重要组成部分,前者用于描述和表达弱点的属性特征,而后者则用于确定计算机和网络系统中由弱点引起的可能的损失量,评估系统的安全风险态势。弱点描述技术是评估技术的基础,而评估技术最终是为了减少甚至消除弱点。然而,日益增长的应用需求对弱点描述和评估技术提出了更高的要求,因此,本文在软件安全弱点的定义范畴内,针对弱点分类、描述语言、主机系统评估和网络系统评估中存在的问题展开了进一步研究。由于已有的弱点分类法普遍缺乏对量化评估以及对揭示弱点关联特性的支持,为此,本文提出了一种支持关联性挖掘的多维量化属性弱点分类法,给出了相应的挖掘算法,并以实验验证了该算法的有效性。该分类法能够支持弱点在特权层面的关联挖掘,从而生成弱点关联图,为弱点的智能检测以及提高弱点评估的准确性提供了帮助。此外,也提供了弱点多侧面细层次上的量化属性,并且突出了弱点对系统机密性、可鉴别性和可用性等安全属性的影响,为量化的弱点描述、评估提供了丰富的信息支持。针对已有弱点描述语言缺乏全面性和通用性等问题,本文提出了一种基于XML的形式化弱点描述语言CVDL,讨论了其应用模式,并以实例详细阐述了弱点的描述和发布。CVDL具有全面性、通用性、以及内容规格严谨性等优点,尤其是能够描述弱点检测、安全影响、属性特征等多方面性质。并且具有良好的结构化描述能力,能够支持弱点信息的收集、交换、整合以及弱点检测、评估或分析等工作的自动化,从而减少了这些后续工作的时间和人力资源的消耗。在主机系统弱点评估中,传统方法往往忽视了若干组件弱点的关联对整个主机系统安全产生的非线性威胁,为此,本文提出了一种基于弱点关联图的通用量化评估方法,并对具有代表性的几款主流操作系统实施了安全评估。该方法引入了弱点关联影响,能够评估弱点关联性对主机系统安全产生更多还原

【Abstract】 With the rapid development of Internet, the Internet-related security events are continually growing. Security issues have become more acute. Theory and practice indicate that why the attack activities like virus, malicious code and network intrusion can badly compromise computer systems is due to the existence of security vulnerabilities in design, implementation and maintenance of computer and software systems. Therefore, to deeply analyze and study vulnerabilities is greatly significant for ensuring computer and network systems.Vulnerability description and assessment technologies are two important parts of vulnerability research. The former is used to describe the attributes and characteristics of vulnerabilities. The latter is used to determine the possible loss caused by vulnerabilities in computer and network systems and evaluate the security risk status of systems. Vulnerability description is the basis of the assessment technology, while the assessment technology is finally to reduce or even eliminate the vulnerabilities. However, increasing application requirements demand description and assessment technologies should be continually developed. So in the conceptual category of software security vulnerability, this dissertation conducts the further research on the problems in vulnerability classification, description language, host assessment and network assessment.Because the existing vulnerability taxonomies at large lack support for the quantitative assessment and uncovering the vulnerability correlations, this dissertation proposes a vulnerability taxonomy for mining the correlations with multidimensional quantitative attributes, gives the corresponding mining algorithm and an experiment is given to validate the effectivity of the algorithm. This taxonomy can support the correlation mining of vulnerabilities to generate a vulnerability relationship graph and provide help for smart identification and promoting the evaluating accuracy. And also this taxonomy gives the quantitative attributes in more sides on finer layers and emphasizes the influences of vulnerabilities on the confidentiality, authentication and availability of systems so that it can provide abundant information for vulnerability description and assessment.更多还原