guardpup

This is our screenshot blocker. Attempt to screenshot will bring this screen up.
This is the social page which shows a user's friends list, close friends list, and analytics. Close friends are able download your images.
This is a post flagged for potential deepfake/malicious content. This disables post abilities and requires further authentication.
This is the vault screen which shows a user's posts and enables to track views as well as downloads.
This is the feed screen which shows posts among a user's followers enabling interactive likes and downloads if given permissions.
This is the AI reverse image search. We return a facial similarity score as well as images at risk for deepfakes.

Inspiration

In response to an exponentially increasing problem of AI deepfake pornography, we need to act NOW to ensure the protection of innocent people online. Unfortunately, as AI becomes more advanced every day, so has the use of AI for malicious purposes. In response to an ever-increasing presence of people "living" on social media and the internet, it is important to give users the tools to protect themselves.

What it does

This software is designed to prevent the spread of manipulated content, safeguard image authenticity, and give users more control over their digital footprint. With features such as deepfake detection search, screenshot blocking, download restrictions, and profile view statistics, users can use our product to limit who has access to their photos on social media and find early signs of post manipulation.

How we built it

Frontend: React & Bootstrap Backend: Node.js & MongoDB AI: Python, Flask & Databricks Open Software Authentication: Auth0

Challenges we ran into

Everyone had to step out of their comfort zone for this project and try new things. For frontend, we had issues learning Bootstrap styling to fit the Auth0 project layout. We felt uncomfortable, but it was required for time management. For backend, building the software for screenshot blocking was a very difficult process. Additionally adapting and learning Auth0's API for authentication was another challenge where we had to adapt and grow. For our AI, we ran into several issues combining multiple models for our deepfake reverse image search. We have had no experience with AI image and face detection prior to this project so learning and deploying our server was possibly the greatest challenge for us.

Accomplishments that we're proud of

Successfully implemented an entire social media website to showcase our security software. Set up authentication and user accounts using Auth0. Deployed a combination of preexisting AI models to track faces, compare them against a database of other images, and return an image similarity score.

What's next for guardpup

Every aspect of our project was built to exist as a security extension to social media websites and other online user platforms. Future implementation would include a broader database for our reverse image search such as Google Images. Additionally, we plan on compressing our project into an SDK file which any platform would be able to use, whether it be a startup or a social media giant.