Apple recently introduced Apple Reference Image, a system designed to cryptographically verify that a photo was taken by a camera rather than generated by AI. Inspired by this concept, two camera enthusiasts developed an open source camera that provides proof of capture at the moment the photo is taken.
Traditional methods focus on detecting AI-generated images after they are created, but these approaches struggle to keep pace with advancing generative models. Instead, this new camera embeds a cryptographic signature directly into the image pixels using steganography, proving authenticity at capture time.
The system hides a signed perceptual hash—a representation of the image’s visual content rather than its exact data—within the image itself. This approach allows the signature to survive common image transformations such as compression and resizing, which typically strip metadata and invalidate traditional signatures.
The signing process is secured by an ATECC608 cryptographic chip, which stores the private key internally and prevents extraction or tampering. This hardware ensures that the signature is trustworthy and cannot be forged.
Apple’s method similarly signs pixels at capture but stores the signature separately in a "digital negative" managed by their Private Cloud Compute system. However, Apple’s approach does not use the open C2PA standard, which is already supported by several major camera and software companies.
While neither solution completely eliminates risks—such as photographing a screen displaying AI-generated content—they represent important steps toward verifying photo authenticity. The open source project is accessible and can be built for under $100, encouraging broader experimentation and transparency in image verification.