OPP – An open protocol for AI image provenance that survives screenshots (github.com)

A new open protocol called OPP has been announced, designed to mint and verify the provenance of AI-generated images, addressing the growing need for transparency in this burgeoning sector. Unlike existing solutions such as C2PA and blockchain networks, which face limitations like metadata stripping on screenshots and impractical gas fees, OPP offers a federated system that survives transformations like cropping and compression. It functions similarly to DNS but for image verification, allowing anyone to mint and verify image signatures, ensuring that AI-generated content can be traced back to its source effortlessly. The technical foundation of OPP is robust, utilizing a tiered verification process that includes a 256-bit SHA-256 hash for exact duplicates, a perceptual hash (PDQ) for near-duplicates, and a CLIP ViT-L/14-based semantic similarity measure. This three-pronged approach allows for lightning-fast verification at scale, with a centralized index powered by a Qdrant vector database capable of handling billions of entries. Open-sourced under the Apache 2.0 license, OPP promotes collaboration and aims to standardize image authenticity in the AI landscape, making it easier for creators and consumers to distinguish genuine AI-generated content from fakes.