Abstract
The existing literature on the connection between attacks and system vulnerabilities often relies on manual techniques. We have developed an approach, called VULDAT, to automatically identify software vulnerabilities and weaknesses from the text of an attack by leveraging the information contained in MITRE repositories and datasets that contain descriptions of attacks and information about attack methods, as well as code snippets that describe the related weaknesses of vulnerabilities. Thus, this research focuses on analyzing attack text descriptions and predicting their effects using natural language processing (NLP) and machine learning techniques. This can be helpful to quickly examine new attacks found in the real world and assist security experts in taking the appropriate actions.