ZK-JPEG: Zero-Knowledge Image Editing and Compression

ZK-JPEG is a cryptographic system that proves an image was correctly produced by compressing a secret, committed input into the JPEG format, enabling robust camera attestation and verifiable edits even after lossy encoding. It targets the problem that digital signatures on raw camera output are invalidated by common downstream operations - lossy compression, selective blurring or redaction - so provenance tools must survive these transformations. ZK-JPEG generates zero-knowledge proofs that a published JPEG was derived from an authenticated source and that permitted transformations were applied, preserving privacy of the original while allowing public verification against a commitment.

The implementation converts Python image-editing code into a zero-knowledge circuit using PicoZK and constructs proofs for the line-point zero-knowledge (LPZK) proof system, demonstrating practical instantiation from off-the-shelf ZK tooling. The design integrates a broad family of image transformations directly into the JPEG compression pipeline with minimal added cost, and emphasizes speed and flexibility suitable for real-world workflows. Authors are Samuel Dittmer, Steve Lu, Kimberlee Model (Stealth Software Technologies, Inc.) and Joseph Near (University of Vermont).