ARVO: Atlas of Reproducible Vulnerabilities for Open-Source Software
Researchers have developed ARVO, a new dataset designed to improve the reproducibility of vulnerability data in open-source software. This dataset addresses the common trade-off between reproducibility, quantity, and diversity in vulnerability datasets by focusing on making each vulnerability consistently rebuildable, triggerable, and analyzable across different versions. ARVO contains over 6,100 real-world vulnerabilities from 311 projects, successfully reproducing 81% of them and achieving 89.4% accuracy in locating corresponding patches. AI