Nonlinear collusion attack on a watermarking scheme for buyer authentication

This paper presents an adaptive collusion attack on a buyer authentication watermarking scheme. To accomplish this attack, the traitors (i.e., dishonest buyers) select the pixels of their watermarked images generated from the same original image and average the selected pixels so as to remove the watermark information. Additionally, the forged image is of higher quality than any watermarked image. Both theoretical and experimental results demonstrate that our attack is very effective.