AppSecBench is an original benchmark of 406 vulnerable/secure code pairs spanning 12 programming
languages, 18 frameworks, 34 vulnerability classes, and 5 difficulty levels. Each record is a
self-contained evaluation case: a vulnerable snippet, its secure counterpart, an exploit sketch,
and the "ground truth" a detector/model is expected to produce (CWE, OWASP, severity, CVSS 3.1,
explainability, fix, and… See the full description on the dataset page: https://huggingface.co/datasets/ismailtasdelen/AppSecBench.
