Software security evaluation using multilevel vulnerability discovery modeling

In this work, we propose a new vulnerability discovery model by predicting the number and probability of occurrence of vulnerabilities of different severity levels in software. The severity prediction assumes that the vulnerability score is a continuous variable distributed over a range of 0-10 as per the widely accepted common vulnerability scoring system. We have further developed a risk assessment model which can be used to define the security level of software and is helpful in risk assessment and patch management. A numerical illustration is done on real-life dataset to validate the proposed model.