Text Clustering on National Vulnerability Database

摘要

In order to solve the problem of taxonomies overlap in Software vulnerability, a method of vulnerability classifying based on text clustering in NVD (National Vulnerability Database) is proposed, and Cluster Overlap Index is used to evaluate Simplekmean, Bisecting KMeans and BatchSom clustering algorithms. 45 main vulnerability Clusters are selected from approximate 40,000 vulnerability records according to Descriptor Dominance Index. These dominant vulnerability taxonomies become our study focuses, which transforms our works from individual to vulnerability taxonomies research.