Applying Deep Learning for Discovery and Analysis of Software Vulnerabilities: A Brief Survey

Abstract

Machine learning has been proved very successful in analyzing and detecting software vulnerabilities. One of the limitations of machine learning is that it heavily depends on human expertise for feature selection. Deep learning can be used to reduce this burden of manual feature selection. The objective of this paper is to survey the feasibility and advantages of applying deep learning techniques for the analysis and detection of software vulnerability. However, deep learning has been developed for a very different class of problems so it needs to be tailored in such a way that it can easily fit in the application of software vulnerability. This paper emphasizes on all such modifications required for deep learning approaches to efficiently accommodate the problem of detection of software vulnerabilities. We have also discussed the various vulnerability databases/resources and some of the recent successful applications of deep learning in predicting vulnerabilities present in the software.