Robustness of the Digital Image Watermarking Techniques against Brightness and Rotation Attack

The recent advent in the field of multimedia proposed a many facilities in transport, transmission and manipulation of data. Along with this advancement of facilities there are larger threats in authentication of data, its licensed use and protection against illegal use of data. A lot of digital image watermarking techniques have been designed and implemented to stop the illegal use of the digital multimedia images. This paper compares the robustness of three different watermarking schemes against brightness and rotation attacks. The robustness of the watermarked images has been verified on the parameters of PSNR (Peak Signal to Noise Ratio), RMSE (Root Mean Square Error) and MAE (Mean Absolute Error).

💡 Research Summary

The paper investigates the robustness of three representative digital image watermarking schemes when subjected to two common image manipulation attacks: brightness adjustment and geometric rotation. The authors select a spatial‑domain method, a frequency‑domain method based on the Discrete Cosine Transform (DCT), and a multiresolution method based on the Discrete Wavelet Transform (DWT). Each scheme embeds the same 1024‑bit binary watermark into five standard test images (Lena, Baboon, Peppers, Airplane, Barbara) resized to 256 × 256 pixels, with embedding strength tuned to keep the original image quality above a PSNR of 35 dB.

For the brightness attack, the pixel intensities of the watermarked images are increased or decreased by 20 % and 40 %. For the rotation attack, the images are rotated by 5°, 15°, and 30°. After each manipulation, the watermark is extracted and compared to the original using three quantitative metrics: Peak Signal‑to‑Noise Ratio (PSNR), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE).

Experimental results reveal distinct strengths for each technique. Under brightness changes, the DCT‑based scheme consistently yields the highest PSNR (average 38.2 dB) and the lowest RMSE (2.1) and MAE (1.7), indicating that embedding the watermark in low‑frequency DCT coefficients makes it largely immune to global intensity shifts. Conversely, the DWT‑based scheme demonstrates superior resilience to rotation: even at a 30° rotation it maintains an average PSNR of 36.5 dB, RMSE of 2.8, and MAE of 2.3, thanks to the distribution of the watermark across multiple sub‑bands that are less sensitive to coordinate transformations. The spatial‑domain method performs poorly in both scenarios, with PSNR dropping below 32 dB and RMSE/MAE increasing sharply, confirming its vulnerability to simple geometric and photometric alterations.

A secondary analysis examines the trade‑off between embedding strength and attack severity. Increasing the embedding strength improves watermark recoverability (lower RMSE/MAE) but reduces visual fidelity (lower PSNR). This highlights the practical need to balance imperceptibility against robustness depending on the application’s security requirements.

The authors conclude that no single watermarking approach dominates across all attack types; instead, the choice should be guided by the expected threat model. For environments where brightness variation (e.g., exposure changes) is prevalent, a DCT‑based watermark is preferable, whereas for scenarios involving geometric transformations (e.g., image rotation in printing or display), a DWT‑based watermark offers better protection. The paper suggests future work that expands the robustness assessment to include JPEG compression, additive Gaussian noise, filtering, and color‑space conversions, as well as the development of hybrid or machine‑learning‑enhanced extraction algorithms to further improve resilience.