Security Patch Detection Repository

Creating the dataset

python data_collection\create_dataset.py --all -o data_collection\data

Notice that this dataset doesn't include the commits' generated data as it is very big. This can be generated by running only :

 python data_collection\create_dataset.py  --commits -data_collection\data

Running the training

python train.py  -a before --model conv1d -k 10 --metadata -c

Citing our work

If you've found our work useful, please cite our paper

@article{farhi2023detecting,
  title={Detecting Security Patches via Behavioral Data in Code Repositories},
  author={Farhi, Nitzan and Koenigstein, Noam and Shavitt, Yuval},
  journal={arXiv preprint arXiv:2302.02112},
  year={2023}
}

Which can be viewed at https://arxiv.org/abs/2302.02112