Combatting Deepfakes: Decentralized Image Authentication (Android)

The Motivation: Why did we build what we built?

In an era where trust and authenticity are paramount in digital media, our project presents a solution at the forefront of blockchain-based image authentication. Every day a total of 24 million AI generated images are created, amassing a total of 15 billion images. The core mission of our project is to be able to distinguish and authenticate raw, unedited images from tampered or artificially generated ones.

Our solution was to build a blockchain-based image authentication application that establishes a NFT collection service tailored for validating images taken on Android devices, generates a unique NFT if an Android image event is emitted, and is resistant to Man-In-The-Middle (MITM) attacks

In the Short Term, we aim to

1. Authenticate/sign images captured by a camera (i.e. Android/Apple) and integrate with existing camera software
2. Prevent signature replication / forgery through immutable metadata unique to individual users
3. Mint NFTs in an aggregated database for all signed, authenticated images
4. Query images against this repository to discover deepfake, edited, or doctored images

However, we do have bold yet achievable Long Term ambitions as well**:**

1. Create a worldwide ecosystem of camera-verified images/NFTs with corresponding image hashes
2. Hardware-level automations: working with chip manufacturers to further secure signature mechanism upon photo capture event
3. Decrease deepfake pervasiveness: Restore consumer trust in digital media

The Nitty Gritty: How does it work?

This is achieved by a cryptographic sign/unsign process initiated only for Android phones that prevents malicious actors from entering false images into a collection of verified images. Our long-term goal is to create a worldwide ecosystem in which it is automatically verifiable whether an image has come from a real camera or it has been faked.

The core principle of our project revolves around the creation of a unique NFT for each authenticated image. Upon detecting new photos within the user's Android phone, our system automatically initiates the authentication process. It waits for an event emitted directly from the Android device, signaling that a photo has been taken. This event-driven approach ensures that the authentication process is triggered only when a genuine photo-capturing event occurs, mitigating the risk of tampering or manipulation by malicious actors.

Our 4-Step Process:

1. NFT Minting:
   1. Automatically mint an NFT on photo/video capture
   2. Store unique hash/NFT in our database as a verified image