Paper 2026/2939

Anonymous C2PA: Feasibility of Large-Scale Anonymous Image Provenance

Kelsey Merrill, Carnegie Mellon University
Sarah Scheffler, Carnegie Mellon University
Abstract

High-quality AI-generated images are easy to create and near-indistinguishable from real photos, heightening the need to reliably establish image authenticity. The Coalition for Content Provenance and Authenticity (C2PA) has developed a technical standard under which cameras and editing software cryptographically sign the images they create, along with metadata and edit history. The standard is widely deployed in products by Google, Sony, Adobe, and others. However, C2PA offers no anonymity for signers. Signed images can be linked to a specific device or person, creating serious surveillance risks. C2PA also lacks formal security definitions, making it difficult for other systems to build on it with provable guarantees. We formalize C2PA v2.5, and then we derive and formalize security and privacy properties desirable for an ideal anonymous image provenance system. Our primary building block for formalization is ring signatures, which let users sign on behalf of a group while staying anonymous within it. We give a new construction of shrinkable ring signatures that supports arbitrary subrings; this can be used to revoke compromised keys by "shrinking'' these keys out of existing signatures. Prior work supported only a small, fixed set of subrings. In our formalization, we find that several desiderata for an anonymous image provenance system are fundamentally incompatible. In particular, we prove that shrinking ring signatures requires work linear in the number of signatures shrunk, and in some settings, linear-sized public evidence; both are impractical at a global scale. We then explore the resulting trade-off space and analyze candidate systems making different choices among the strength of authenticity guarantees, network access requirements, and post-quantum security. Our work answers calls for a more private image provenance system and identifies several paths toward practical solutions.

Metadata
Available format(s)
PDF
Category
Applications
Publication info
Preprint.
Keywords
C2PAimage provenancering signaturesshrinkable ring signaturesanonymity
Contact author(s)
kmerrill @ andrew cmu edu
sscheffl @ andrew cmu edu
History
2026-11-22: approved
2026-11-18: received
See all versions
Short URL
https://ia.cr/2026/2939
License
Creative Commons Attribution
CC BY

BibTeX

@misc{cryptoeprint:2026/2939,
      author = {Kelsey Merrill and Sarah Scheffler},
      title = {Anonymous {C2PA}: Feasibility of Large-Scale Anonymous Image Provenance},
      howpublished = {Cryptology {ePrint} Archive, Paper 2026/2939},
      year = {2026},
      url = {https://eprint.iacr.org/2026/2939}
}
Note: In order to protect the privacy of readers, eprint.iacr.org does not use cookies or embedded third party content.