| Title |
Automatic Detection of Software Security Vulnerabilities |
| Authors |
김성민(Sung-Min Kim) ; 김동관(Dong Kwan Kim) |
| DOI |
https://doi.org/10.5370/KIEEP.2021.70.3.157 |
| Keywords |
Security Vulnerability; Deep Learning; Common Vulnerabilities and Exposures (CVE); Code Changes |
| Abstract |
Software vulnerability refers to the characteristic that software can be exploited by attackers. Unauthorized actions can cause economic loss or damage to human life. Therefore, security vulnerabilities should be managed to prevent a malfunction of a software system. This paper provides a deep learning-based system that automatically detects software security vulnerabilities. The proposed detection system builds datasets with vulnerable and non-vulnerable functions for a supervised learning model. These datasets are collected from the CVE databases and GitHub repositories. The automation detection model achieved a high f1-score of 98%. Furthermore, the proposed model showed better classification performance than traditonal machine learning algorithms |