.png){kind=icon}
.png){kind=icon}
Return
Vulnerability identification by harnessing inter-connected multi-source information
.png){kind=icon}
PRE
AI
L
H
X
L
Y
Y
DOI：10.1016/j.jss.2026.113001.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• We propose VPFinder, a deep learning based method that identifies vulnerabilities by harnessing interconnected multi-source information. By learning and fusing the high-level semantic information from bug reports, commit messages and patches. VPFinder could effectively recognize vulnerabilities and their corresponding types. • We evaluate VPFinder and the experimental results show that VPFinder achieves an F1 score of 0.941 in vulnerability prediction and 0.610 in vulnerability type prediction, outperforming state-of-the-art approaches. • We make our dataset, tool, and model publicly accessible at: https://anonymous.4open.science/r/VPFinder-5CE4
Journal
IF:
4.1
Papers:
5.4K
Citations:
8.4K
Organization
Cited Papers