Vulnerable Source Code Collected from Open Source Repositories for Dataset Generation

This repository holds five datasets that list existing Buffer Overflow vulnerabilities in more than 10000 source code files written in C. Particularly, it is suitable for extracting features and creating training datasets for Machine Learning. The data was collected from open source repositories linked to SonarCloud using SVCP4C on different dates.

Datasets

Name	Total Files	Checksum
DataSet_07032019.tar.gz	2305	abc7e173fca5d1e7b22313dfade2be19297d6e6735e4d325301a2f410f488797
DataSet_03062019.tar.gz	2378	ba367f4c4c21e26f6de79652185758394d80998d88b14a0bcbecc299b6336a3d
DataSet_16082019.tar.gz	2262	c12599c8412629925f09821bdf2a74970afd631a9e376cd7fd67bdc5fef9ec3f
DataSet_20082019.tar.gz	2258	6c69f14cf6e839955d97e48afe502914715d84d409170ea0ba499107ec902943
DataSet_31082019.tar.gz	2257	b3984fe3d91426b607f89222d44fef7bfcbb13af894ea2f29afadf96365da1de

Vulnerabilities format

Vulnerabilities are listed using comments appended at the end of each file (see SVCP4C repository). Such comments follow the format // starting_line,starting_offset;ending_line,ending_offset (with offset being the column). For example, file DataSet_03062019.tar.gz/bzip2debianstretche1.0.6/bzip2.c has four vulnerabilities tagged like:

//						↓↓↓VULNERABLE LINES↓↓↓

// 1126,3;1126,9

// 1153,9;1153,15

// 1341,9;1341,15

// 1734,6;1734,12

Reference

For scientific publications, please reference this repository using:

Plain

Raducu, R., Esteban, G., Rodríguez Lera, F. J., & Fernández, C. (2020). Collecting Vulnerable Source Code from Open-Source Repositories for Dataset Generation. Applied Sciences, 10 (4), 1270. DOI: https://doi.org/10.3390/app10041270

BibTeX

@ARTICLE{Raducu2020,
  Title     = {Collecting Vulnerable Source Code from Open-Source Repositories for Dataset Generation},
  Author    = {Raducu, Razvan and Esteban, Gonzalo and Rodr{\'\i}guez Lera, Francisco Javier and Fern{\'a}ndez, Camino},
  Journal   = {Applied Sciences},
  Volume    = {10},
  Number    = {4},
  Pages     = {1270},
  Year      = {2020},
  Publisher = {Multidisciplinary Digital Publishing Institute},
  Doi       = {https://doi.org/10.3390/app10041270}
}

License

Datasets are licensed under CC BY-ND 4.0.