Yours, Truly

Your Digital Truth Companion. Use Cryptographic protocols to secure your data. Browse authentic moments and detect Al content with confidence. Celebrate genuine human creativity. Reclaim what's real.

Comment

Inspiration

We've noticed something concerning while scrolling through Instagram, TikTok, and YouTube lately. There's a ton of fake AI-generated content out there with made-up claims and stories.

You've probably seen it too - deepfakes of celebrities or those ridiculously over-the-top videos designed just to get you to comment or share (helping those fake accounts rack up views).

What we really want is a way to protect genuine creators from having their content stolen. We don't want someone taking a real person's face or voice and twisting it into something the original creator never said or did. For those of us who actually want to enjoy authentic human content, we need a better way to verify what's real and what's not...

What it does

Yours, Truly adds a digital signature to every picture or video you take with our app using cryptographic protocols (C2PA). This signature stays in your content's metadata, so anyone can trace it back to you as the original creator.

We also use smart contracts to keep track of who edits your photos and who doesn't. Our map feature helps you see where your content travels after you post it online.

We've built in AI detection tools too - we use a ViT (Vision Transformer) model to analyze images and store those results in our database. When you browse our feed, we prioritize showing content tagged as human-created.

That said, we're not anti-AI at all! We're just using technology to help track what's original versus copied content, making sure you have control over your privacy as the original creator.

How we built it

For our app's design and user interface, we built everything using Swift and SwiftUI, which are Apple's programming languages for creating iOS apps.

On the backend, we're using a serverless framework on Google Cloud Platform (GCP). We host our AI model on Cloud Run, which automatically scales as needed.

Every time you upload or take a picture, we use Firebase triggers to call our AI model. The model then analyzes your image and sends back predictions. This all happens seamlessly in the background while you're using the app.

Once your picture has been captured or uploaded, a cryptographic manifest is generated and embedded in the metadata through the open-source Rust C2PA Foreign Function Interface. Following the C2PA standard, this manifest is signed by C2PA-trusted certificates and contains provenance information such as content authority, timestamp, and location generated to allow traces back to the original user. If the photo has been edited in any way, it is simple to check the manifest signature and the manifest itself to confirm if the integrity was changed and which specific edits were made through our watermarking feature on the tools page.

Challenges we ran into

1) The first challenge was SwiftUI had underwent some big changes since the last time we used it. So that was a learning experience. 2) We were not sure if we wanted to use the properties of a camera (i.e color filter array) to uniquely identify the user. We also did not know which protocol to use first. It took us a lot of research to decide between a watermarking vs camera properties or C2PA. However, after weighing the pros and cons we decided that since various large companies support C2PA there will be a push to improve and have it implemented widely and systematically. 3) Using Rust with Swift was a bit tricky and meticulous since it was our first time but we figured it out!

Accomplishments that we're proud of

We are proud of having a fully functional application which integrates the C2PA standard into both user-generated images through the camera interface and user-uploaded images through the post interface. Yours, Truly secures the integrity of human-generated images through creating a manifest signed by C2PA-trusted certificate authorities, which allows users to reclaim control of where their data is being sent over its lifecycle. Our AI model has 97% masked accuracy.

What we learned

We learned that there are big companies that are trying to figure out this issue. There are research groups at MIT, deep mind, Meta, etc that are all trying to figure out what the best solution entails to secure original Human data, whether that be watermarking, c2pa, etc. It was actually amazing to see how research papers were trying to be implemented and really inspired my teammate and I!

We learned about cloud + mobile i.e firebase triggers and connecting github repos to cloud which will run anytime you push a new change which were all really cool to implement in our app!

What's next for Yours, Truly

We would like to seek funding for Yours, Truly. There is not a lot of competition in this market and recently one main competitor is starting to gain traction by its partnership with Meta to have the first "verified" election content. By continuing to build features for Yours, Truly we can be a competitive alternative. As for actual social media apps which promote AI-generated content, we can be an alternative for users who value real human connections through social media. We plan to conduct more research on augmentation of the currently employed cryptographic protocol as well as review user feedback on the issue of AI content consumption and intellectual data privacy. With this, we can continue to make Yours, Truly a successful platform for users who want to reclaim what's truly yours.

Built With

Share this project:

Updates