SQLidify

Logo
Our Mission
Output

Inspiration

The best way to learn to code is usually through trial and error. As a team, we all know first hand how hard it can be to maintain the proper standards, techniques, and security practices necessary to keep your applications secure. SQLidify is a teaching tool and a security tool all in one, with the goal of helping coders keep their applications secure.

What it does

SQLidify uses our own unique dataset/training model which consists of over 250 labelled data entries to identify SQL vulnerabilities in an application. To use it, simply paste your code into our website where our machine learning model will identify vulnerabilities in your back-end code, and then will suggest strategies to fix these issues.

How we built it

We used a Flask, python based backend that handles API calls from a front end designed in React.js and Tailwind CSS. When called, our python backend reads data from users and then sends the data to our AI model. At the same time, our own simplified natural language processing model identifies keywords in specific lines of code and sends these lines individually to our AI model. The model makes a prediction for each which is then compared to help improve reliability. If the predictions don't match, further instructions are sent to the user in order to improve our accuracy.

The AI is designed using Cohere's classification workflow. We generated over 250 code snippets labeled as either vulnerable or safe. We have another model that is triggered if the code is determined to be vulnerable, which will then generate 3 appropriate options to resolve the vulnerabilities.

Challenges we ran into

We had trouble setting up cohere and getting it to integrate with our application, but we were luckily able to resolve the issues in time to build our app. We also had a lot of trouble finding a dataset fit for our needs so we ended up creating our own from scratch.