Enhanced Categorization of Cybersecurity Vulnerabilities

According to the public National Vulnerability Database, maintained by the National Institute of Standards and Technology, the number of registered cybersecurity vulnerabilities keeps growing despite all known techniques for developing secure software and detecting vulnerabilities. Categorizing cybersecurity vulnerabilities is essential for choosing effective protection mechanisms to prevent or mitigate vulnerabilities in software and reduce cyber risks. In this paper, we present a machine learning-based methodology to categorize cybersecurity vulnerabilities using their descriptions registered in the public National Vulnerability Database. We will show that handling imbalanced data with Synthetic Minority Over-sampling Technique improves the categorization model accuracy by $20 \%$.