Are you an LLM? Read llms.txt for a summary of the docs, or llms-full.txt for the full context.
Skip to content

Landscape

Several technologies exist for image authenticity and provenance. Most solve for different problems, such as edit history, AI detection, trust scoring, rather than proving a photo was authentically captured on a real device.

ZCAM is a complete solution for mobile photo authenticity. It combines hardware-backed attestation, optional zero-knowledge proofs, and embedded verification into a single system designed for human-taken photos on mobile devices.

Comparison

ApproachHardware AttestationZK ProofsEmbedded VerificationUser-friendly OutputCamera PhotosUse Case
ZCAMComplete mobile photo authenticity solution
C2PAEdit history and provenance
Physical CamerasHardware signed capture with limited support
JPEG TrustTrust evaluation framework
SynthIDAI watermarking and detection
AI Content MarkingLabeling AI-generated content

Other Approaches

C2PA

C2PA is an open standard for creating a verifiable history of an image. It tracks provenance but does not necessarily prove authenticity. ZCAM uses C2PA as its container format while adding hardware-backed guarantees that transform it from a provenance system into an authenticity system.

For a deeper dive into how C2PA works and how ZCAM extends it, see C2PA.

JPEG Trust

JPEG Trust is a framework for evaluating image trustworthiness. It works on top of C2PA, using manifest data to generate trust scores. It is an evaluation layer, not a capture or verification system, and could work alongside ZCAM-generated manifests.

SynthID

SynthID, developed by Google, embeds invisible watermarks into AI-generated images. Unlike metadata-based approaches, these watermarks survive cropping and editing. A separate model analyzes images to determine whether they're AI-generated. SynthID focuses on AI detection with probabilistic results. ZCAM focuses on authenticity with cryptographic guarantees.

AI Content Marking

Several companies provide solutions for marking AI-generated content, using different approaches:

  • Meta adds durable watermarks specifically to AI-generated content.
  • OpenAI generates C2PA-compatible images with DALL-E and ChatGPT.

These solutions label AI content, whereas ZCAM verifies human-captured photos.

Hardware Camera Support

We are beginning to see camera companies roll out C2PA-supported cameras. Leica was the first camera company to really lean into C2PA starting in late 2023, with other companies adding support since then.

It's worth noting that different companies have different levels of integration and security. For example, Leica uses a hardware-backed signing system to sign the manifests, whereas other camera companies use a software-based solution. Obviously, the hardware-backed solutions will be harder to exploit or generate signatures over non-authentic pictures.

Each camera company only has C2PA support for a subset of their camera models. Generally speaking, their flagships seem most likely to have C2PA support.

BrandSupported Models
LeicaM11-P, M11-D, SL3-S
SonyAlpha 1, 1 II, 7 IV, 7S III, 9 III (plus video on FX3/FX30/PXW-Z300)
CanonEOS R1, EOS R5 Mark II
FujifilmGFX100S II, X-T50
NikonZ6 III (firmware + cloud cert)

For more information, see: Yawnbox's blog post.

When to Use ZCAM

ZCAM is the right choice when you need:

  • Hardware-backed proof that a photo was taken on a real device
  • Cryptographic verification rather than probabilistic detection
  • A mobile-first solution

Other approaches may be more optimized for marking AI-generated content (SynthID, AI Content Marking), tracking edit history (C2PA), or non-mobile platforms.