Jack-a-test

Inspiration

Having a secure codebase that does not expose any vulnerabilities has never been more important than now, especially with the increasing risk of other agents (who have a variety of capabilities) who are trying to take advantage of such vulnerabilities.

What it does

Enter a link of a repository written in the jac programming language. A custom trained model (5k code snippets as dataset) analysis the repository and gives a score of 1-100. It also finds the specific code snippets and points out what kind of security vulnerability it has

How we built it

We used jac for the overall architecture of the software i.e the frontend and sending the repo to the model. The code for the model was written in python.

Challenges we ran into

We ran into a challenge of creating a curated dataset and also training the model in the first place.

Accomplishments that we're proud of

We are proud of having a model that can accurately point out security vulnerabilities in a repository written in the jac programming language

What we learned

We learned that spending time on developing a curated dataset clearly improves accuracy of your model

What's next for Jack-a-test

Get feedback from Jac and improve based on that

Built With

• hugging-face
• jac
• python