Attack operators for digitally watermarked images

Experiments are reported in which the performance under malicious attack of discrete cosine transform-based digital watermarking has been investigated. Image-processing operators (attack operators) have been found that are able to impair significantly the embedded watermark without any knowledge of the secret keys used for the embedding process. Of these, the Laplacian removal operator proved to be particularly effective. Such operators need to be recognised and understood in the design of effective watermarking schemes