Machine Learning for Automated Vulnerability Detection in Source Code

Masters Project at University of Bristol

All of our code can be found in the code folder. Files are individually documented, while the README.md file guides you through the narrative and ordering of the project. The data used in this project can be found in data. Generated figures and images can be found in images.

Most of the code should run in our Docker environment (see `Development" below for setup details), with the exception of:

1. Progol :: There is a working Progol in the Docker image (available in the /progol/ directory). However, to repeat our results, our scripts should be ran on BlueCrystal using CProgol 4.4 compiled with GCC 5.4.
2. Aleph :: We do not have redistribution rights for the ILP system we used.

Development

We use git-lfs for storing larger files (primarily those in the ./data/ and ./images/ directories). This means that you need git-lfs installed before running git clone. See this guide for more info. If you get errors loading csv files into Pandas, it's probably a problem with git-lfs. Further, if there are errors when git-lfs is pulling files, it might be because we've ran out of bandwidth on this account, please get in touch with Dan (dj311) and he'll fix it.

Use Docker to get a working development environment. The easiest way is to pull down the latest build from docker hub via:

docker pull djwj/uob-summer-project

Once the image is pulled, run the container in the directory where you've cloned this repo, via:

• docker run --rm --name uob-summer-project --volume $PWD:/project --publish 8888:8888 --interactive --tty djwj/uob-summer-project

The above command will start a Jupyter server. If you need a shell, you can exec into the container with docker exec -it uob-summer-project /bin/bash. That's an easy way to run Prolog or similar.

Adding New Dependencies

First, add them to the right place depending on their type:

• Python dependencies can be added to the requirements.txt.
• Add native libraries via the apt install ... line in the Dockerfile.

Next, the Docker image needs to be rebuilt, then pushed up to the docker hub. Do this via:

docker build --tag djwj/uob-summer-project .
docker push djwj/uob-summer-project