ePrint 2026/2039 Defines ZK-JPEG Circuit for Committed JPEG Compression

Paper 2026/2039 constructs a zero-knowledge circuit for the full JPEG pipeline. Authors convert Python editing scripts through PicoZK into line-point zero-knowledge proofs. The system accepts a committed raw image, applies lossy DCT quantization and Huffman coding, then outputs a proof that the published JPEG matches the committed source under the declared transformations. Prior ZK image schemes failed after quantization; ZK-JPEG folds quantization tables directly into the circuit. Benchmarks in the paper show proof generation under 12 seconds for 1024x1024 images on commodity CPUs when using standard quality tables. Integration cost for redaction or blur operations adds fewer than 15 percent overhead because the same circuit already materializes the DCT coefficients. The construction therefore survives the exact transformations that invalidate camera signatures under C2PA and similar attestation frameworks. The paper omits direct comparison against SNARK-based alternatives such as those built on Groth16 or Plonk for image pipelines. It also provides no side-channel measurements for the committed input phase. Operationally this means camera vendors can embed the prover in firmware only after additional leakage analysis and after standardization of the commitment format. Next steps include porting the PicoZK output to hardware accelerators already deployed for HEIF and AVIF pipelines. Regulatory pressure from deepfake disclosure rules will likely force at least one major sensor vendor to publish a reference implementation within 24 months.