Authenticity Witness

Motivation:

We hate seeing misinformation, stupidly AI-generated images and fake photos. There's lots of this stuff on Social Media and in news, unfortunately.

The easily AI-created content makes internet a big trash bin with fake-personalized content. It's reducing information value and gives an opportunity to manipulate masses.

We decided to focus on photography. What is the way to make sure that the photo you see is taken by a human, not tampered, and not generated by AI?

Solution:

TLDR:

• Let the orb-verified user take a picture in World Mini App (Thanks Worldcoin)
• Hash the picture with a fancy compression-resistant algorithm
• Put the image on IPFS, making it accessible to everyone
• Attest the hash, IPFS link, and metadata on the blockchain (Thanks Sign Protocol)
• If the picture is still fake (picture of a picture or another uncontrollable attack), let the DAO do the job of banning the user. (Thanks Polygon, Chainlink)

The solution is to make sure that from the moment of creation of the picture up to someone's observation of this picture, the photo is not tampered or altered. We approached the problem using web3 technologies.

All magic happens inside a Mini App powered by World. At first, to use the app, you have to get your World ID orb-verified, this is an important step, allowing Authenticity Witness to make sure any participant of the social network is a real person and is unique.

In the app you can take a picture, and ask the app to attest it and post a high certainty proof that your picture is real. The app communicates with the OS of your mobile phone to take a picture and instantly hash it with Wavelet Hashing algorithm and put it in IPFS. In short, this allows to make the smallest modification to the picture, and the hash will stay the same (A good thing for journalists and photographers, who like filters).

Then this hash, IPFS link, as well as the metadata of the picture are attested by the Witness DAO on the Polygon blockchain using Sign Protocol. These attestations are public so anyone can check the authenticity of the picture, the author has just to attach the link of the attestation to the picture.

The DAO is responsible for checking the authenticity of the picture. By default, as the picture is attested, it is considered real. But if someone finds a way to cheat the system, the DAO will ban the user and the picture will be marked as fake. And thanks to World ID, the author (human behind the author) will be banned permanently from the network. So he'd better not try to cheat.

Behind the Scenes We have Chainlink, helping to update the vote decisions with automated calls, AKA Upkeeps. It's calling the essential for votes endVotes() function, which may ban users permanently from attesting their fake images.