摘要
为解决现有软件漏洞分类重叠性和实用性低等问题,提出了在漏洞实例聚类基础上的漏洞分类方法。对漏洞数据库(national vulnerability database,NVD)的漏洞描述字段进行文本聚类,并且使用聚类重叠性指标评估Simplekmean、BisectingKMeans和BatchSom聚类算法的效果,依据领域主导度选择典型的漏洞类型。实验结果显示近NVD中四万条漏洞数据聚类成45类典型漏洞,从而使软件漏洞研究工作从个体研究转变成对主导漏洞类型的研究。
In order to solve the problem of overlap and low efficiency in software vulnerability taxonomies,proposed vulnerability classifying method based on text clustering of vulnerability descriptor fields in NVD（national vulnerability database）,and used cluster overlap index to evaluate the performance of Simplekmean,BisectingKMeans and BatchSom clustering algorithms.The experimental results demonstrate that 45 dominant clusters are selected from approximate 40 000 vulnerability records in NVD according to descriptor dominance index,and it transforms the vulnerabilities research focuses from individuals to vulnerability taxonomies.
出处
《计算机应用研究》
CSCD
北大核心
2010年第7期2670-2673,共4页
Application Research of Computers
关键词
漏洞数据库
文本聚类
聚类重叠指标
主导漏洞类型
vulnerability database
text clustering
cluster overlap index
dominant vulnerability taxonomies
