A Survey of the Software Vulnerability Discovery Using Machine Learning Techniques.

Nowadays, the study of vulnerability discovery has been attracted the widespread attention and the experts have proposed many different approaches in the past decades. To optimize the efficiency of the method, machine learning techniques are introduced into this area. In this paper, we provide an extensive review of the work in the field of software vulnerability discovery that utilize machine learning techniques. For the three key technologies of static analysis, symbolic execution and fuzzing in vulnerability discovery field, we first explain the basic principles respectively. Afterward, we review the research situation of software vulnerability discovery using machine learning techniques. Finally, we discuss both advantages and limitations of the approaches reviewed in the paper, and point out challenges and some uncharted territories in the three categories. In this paper, a brief study of the software vulnerability discovery using machine learning techniques is given, which is helpful to carry out the follow-up research work.