This repository contains sample code to reproduce the research done for the bachelor thesis "Deep Learning-Based Code Vulnerability Detection: A New Perspective" at SAP Security Research.
The repository implements an GNN evaluation pipeline including cross-validation as well as pretraining schedules.
To run the experiments, the DiversVul dataset (Chen, Yizheng, et al. 2023) must be downloaded, graphs need to be parsed with the cpg tool and python packages in 0_install are required. Further, scripts in codegraphs/diversevul/ produce intermediate pickle files for cross-validation and filtering large and small graphs, which CodeGraphDataset.py requires to load the datasets.
All configuration files can be found in configs/. By switching out the filename in 1_train.py different models can be run. 2_helper_get_best_run.py summarizes results from cross-validation.
- The main test results are produced with the
configs/7_*andconfigs/9_*files. - Visualizations from the paper are made with scripts in
utils/ - Different models as well as the training script are specified in
models.
No known issues.
Create an issue in this repository if you find a bug or have questions about the content.
For additional support, ask a question in SAP Community.
If you wish to contribute code, offer fixes or improvements, please send a pull request. Due to legal reasons, contributors will be asked to accept a DCO when they create the first pull request to this project. This happens in an automated fashion during the submission process. SAP uses the standard DCO text of the Linux Foundation.
Copyright (c) 2024 SAP SE or an SAP affiliate company. All rights reserved. This project is licensed under the Apache Software License, version 2.0 except as noted otherwise in the LICENSE file.
