Abstract
We present a new approach for modeling color attacks of rgb-color watermarked images. We have used a based interpolation watermarking algorithm and supposed that the attacks are simulated by a scaling of the colors followed by a translation. We give bounds for the extracted watermark depending on the original image, the watermark and the attack. Different attacks like LSB, embedding another watermark, Stirmark have been simulated and the quality of the extracted watermark is discussed in each case.