Abstract
This paper is intended to demonstrate that the widely researched adaptive watermarking is noninvertible in nature if it can induce and exploit the perturbation in an attacker's reverse engineering properly. The intrinsic noninvertibility of some adaptive watermarking is disclosed and evaluated. Under the technologies that exploit it, attackers have great difficulty dividing a released version into their claimed original and scaled watermark, and in the meantime making the latter be the particular adaptive result based on the former. Their reversed solutions are drastically perturbed and perceptually unacceptable. Compared with the existing noninvertible method based on cryptography, the proposed method is more efficient because it disposes of the cryptographic processing. Furthermore, it also shows a promising way of making a public watermarking scheme, which is considered intrinsically invertible, noninvertible.