Page 22 - Detecting deepfakes and generative AI: Report on standards for AI watermarking and multimedia authenticity workshop

Detecting deepfakes and generative AI: Report on standards for AI
watermarking and multimedia authenticity workshop

Figure 9: Mosaic of AI-generated images shown during the workshop by Andrew
Jenks to illustrate the rapid advancement of AI in contrast with Figure 3� Warning:
these images are not real�

The main advantage of Content Credentials is that they can combine secure metadata,
watermarks, and content fingerprinting to provide a solution to establish the provenance of
multimedia content.

Some of the techniques for recording and preserving provenance data include:
• Secure metadata: This is verifiable information about how content was made associated
with the content itself, in a way that cannot be altered without leaving evidence of
alteration. Content Credentials can provide information about the provenance of any
piece of multimedia content or composite. It indicates whether a video, image, or sound
file was created with AI or captured in the real world with a device like a camera or
audio recorder. Content Credentials are designed to be chained together to provide
information about how content may have been altered, what was combined to produce
the final content, and even what device or software was involved in each stage. The
various elements of provenance in the lifecycle of a digital asset can be combined in
ways that preserve privacy and enable creators and consumers to obtain information
on the origins of the digital asset and how it was created. The metadata can be either
signed using digital signatures or preserved using distributed ledger technologies such
as blockchain to maintain the integrity of the provenance data and verify its authenticity.
• Watermarking: Watermarking is a technique to embed hidden information that cannot
be seen by people. It can be decoded using a watermark detector. State-of-the-art
watermarks can be impervious to alterations such as the cropping or rotating of images or
the addition of noise to video and audio. Importantly, the strength of a watermark is that
it can survive efforts to remove secure metadata such as taking screenshots or pictures of
pictures, or re-recording audiovisual content.
• Fingerprinting: This refers to a way to create a unique code or hash based on pixels,
frames, or audio waveforms that can be computed and matched against other instances
of the same content, even if there has been some degree of alteration. The fingerprint
can be stored separately from the content as part of Content Credentials. When someone
encounters the content, the fingerprint can be re-computed and matched against a
database of Content Credentials and its associated stored fingerprints. The advantage

14