It is shown in this paper that, even with a non-invertible watermarking algorithm or an asymmetric watermarking protocol, it is still possible to effect an invertibility attack, which relies on a forged watermarking algorithm, a counterfeit mark, and a fake key. Two examples are given to show the vulnerability of the unfortified non-invertible algorithm/asymmetric protocol. As a solution, a secure watermarking protocol is proposed, which establishes correlation between the watermarking algorithm and the embedded mark. © 2004 Elsevier B.V. All rights reserved.
Zhang, X., & Wang, S. (2004). Invertibility attack against watermarking based on forged algorithm and a countermeasure. Pattern Recognition Letters, 25(8), 967–973. https://doi.org/10.1016/j.patrec.2004.02.007