Survey: Machine Learning Algorithm Efficacy Static Software Analysis

In this day and age, there is a constant growth in technology and a flood of software and devices in the market. With this comes the need for security improvements. Software analysis alone can take substantial time, cost, and extraordinary talent. There is also a large repository of shared code available. Social coding is an avenue that plays into the reason of having a way to detect vulnerabilities, be it originally in the code or added into later, is even more of a concern. One possible way to assist in the process of vulnerability detection is the use of machine learning. Machine learning is something that has proved to be efficient, cost-effective, and beneficial so far in this aspect. With the use of static analysis, we think it is the future for software developers and analyzers. It is important to discuss where we are now with utilizing machine learning and where we can go. This paper provides the foundation to begin this discussion by developing an understanding of how machine learning algorithms are being used to detect vulnerabilities in software and their limitations.