Watermarking’s Major Shortcomings.

ALL Invisible & Vulnerable !!

Watermarking has emerged as one of the more promising strategies to identify AI-generated images and text.

Just as physical watermarks are embedded on paper money and stamps to prove authenticity, digital watermarks are meant to trace the origins of images and text online, helping humans spot deepfaked videos and bot authored books etc.

“We don’t have any reliable watermarking at this point.”

“We broke all of them.” S. Feizi.

Feizi and his research team looked at how easy it is for bad actors to evade watermarking attempts.

He calls it “washing out” the watermark.

To demonstrate how attackers might remove watermarks, the study shows how it’s possible to add watermarks to human generated images, triggering false positives.

Some researchers have claimed that visible = high perturbation watermarks might be developed to withstand attacks.

Feizi and team’s recently found that even the most promising type can be manipulated.

“Perhaps we should get used to the fact that we are not going to be able to reliably flag AI-generated images.” Feizi.

Ben Colman,CEO of the AI-detection firm Reality Defender agrees:

“Watermarking at first sounds like a noble and promising solution, but its real-world applications fail from the onset when they can be easily faked, removed, or ignored.”

The dilemma in AI watermarks has even created a new business:

The quick watermark-removal service.

So it might be time to revalue the expectations placed on watermarking AI ?

T.Goldstein sees watermarking as a form of harm reduction and useful for catching lower-level attempts at AI fakery, even if it can’t prevent high-level attacks.

The Fully Robust Watermark is still to be developed, if that’s the sollution at all?

It’s a Real Time: Cat Catching The Mouse Game out there.

By My