Investigating the possibility of increasing the resilience to attack of the differential energy watermarking embedding method

Abstract

The aim of this paper is to investigate the possibility of increasing the robustness of digital watermark embedding by the differential energy watermarking (DEW) embedding method to static attacks: median filter, Gaussian filter, threshold rounding, noise reduction, compression; to geometric attacks: 90-degree rotation, slight clipping, scaling; attack on rewrite. The peculiarities of digital watermark embedding by DEW method and ways to improve the reliability of this method using Watson’s visual model are discussed. One of the main problems in digital watermark embedding by the DEW method is that it embeds information in high-frequency coefficients that are discarded by filters, and the DEW method does not take into account the effect that changes in discrete cosine transform (DCT) coefficients have on the original image. As a solution to these problems a modification of the method based on the Watson visual model has been proposed. The model estimates the maximum allowable value that can be changed without visual influence. Thus, by matching the discrete cosine transform coefficients in the DEW algorithm to the values of frequency sensitivity of Watson’s visual model, one refuses to embed the watermark in the area with complex textures in order to reduce the impact on the visual quality and filtering effects. In this paper the changes introduced into the image by digital watermark embedding were evaluated using structural similarity (SSIM), and the stability of the digital watermark was evaluated using Pearson coefficient. Based on the data presented in the paper, a comparison of DEW method and DEW method using Watson’s model was carried out. According to this study, the DEW method using Watson’s model is resistant to repeated recording attacks, partially resistant to geometric and statistical attacks, while the DEW method is not resistant to static and geometric attacks.

Keywords

• digital watermark, discrete cosine transform, dew method, watson model, pearson coefficient, ssim.