The Effect of Topic Modelling on Prediction of Criticality Levels of Software Vulnerabilities.

In this day and age, software is an indispensable part of our per diem endeavours, thereby keeping a check on exploitable vulnerabilities has become a vital function of a software firm. The motivation of this paper is to have better understanding of vulnerabilities, creating a tool for the industry practitioners to identify a critical vulnerability that could be detrimental for the firm’s assets. In this article, 1999 vulnerabilities related to Google Chrome was analysed to understand the behaviour of vulnerabilities. The identification of trends and patterns using topic modelling technique lead to extraction of topics. The extricated topics were then implemented in 10 classifiers to foresee the criticality of the vulnerability. The resulting performances were also assessed with the classifiers without implementing topic modelling techniques. A 10-fold validation was conducted on the suggested prediction model.