PrimeVul
收藏资源简介:
PrimeVul数据集是一个为C/C++代码漏洞检测而设计的数据集,用于在现实场景下训练和评估代码语言模型。该数据集提供了预分割的训练集、验证集和测试集,包含了大约7000个漏洞函数和229000个良性函数,覆盖了超过140个常见弱点枚举(CWEs)。数据集注重准确的标签、最小的数据污染和丰富的元数据,以支持高级分析。各个数据集分割部分包括源代码、二进制标签以及提交、漏洞和文件级别的元数据。
The PrimeVul dataset is a specialized dataset designed for C/C++ code vulnerability detection, which is used to train and evaluate code language models in real-world scenarios. This dataset provides pre-split training, validation, and test sets, containing approximately 7,000 vulnerable functions and 229,000 benign functions, covering more than 140 Common Weakness Enumerations (CWEs). The dataset prioritizes accurate labeling, minimal data contamination, and rich metadata to support advanced analysis. Each dataset split includes source code, binary labels, as well as metadata at the commit, vulnerability, and file levels.




