Hello Adam Rose and Bots Don't Cry,
I just saw this thread.
C2PA is an Adobe-centric solution that does not validate content, metadata, or signatures. Because it is based on "trust", it does nothing to prevent forgeries or false attribution.
I just saw this thread.
C2PA is an Adobe-centric solution that does not validate content, metadata, or signatures. Because it is based on "trust", it does nothing to prevent forgeries or false attribution.
Comments
https://spectrum.ieee.org/meta-ai-watermarks Article says Meta's AI Watermarking, but talks about C2PA's approach. "Flimsy, at best".
https://www.technologyreview.com/2023/07/31/1076965/ MIT Tech review says C2PA will "not stem the harm of machine-generated misinformation."
https://hackaday.com/2023/11/30/falsified-photos-fooling-adobes-cryptographically-signed-metadata/. Hackaday describes how to use Adobe's C2PA solution to create authenticated forgeries.
https://hackerfactor.com/blog/index.php?/archives/919-Closed-Standards.html