abstract.tex

\subsection*{Abstract}

As the number of softwares being produced and the size of each software grow, vulnerabilities in softwares are also becoming more prevalent.
In order to detect these vulnerabilities, efficient detection techniques are required.
How to detect vulnerabilities has been worked on for a long time and various approaches have been introduced, for instance, static or dynamic program analysis.
Although they had achieved their own success, they are still practically infeasible to be applied to complex modern softwares.
Various requirements such as modest execution time, accuracy and granularity should be met to efficiently detect vulnerabilities.
To achieve such objectives, I propose vulnerability detection method combining comprehensive representation and deep learning.
I use code property graph as such representation, and compare various deep learning models to reach my objectives.