Enhancing the Robustness of the E_BLIND/D_LC Digital Watermarking Algorithm Against DCT-Filtering Removal Attacks

The paper focuses on evaluating and improving the robustness of watermarking algorithms against DCT-filtering, a common image processing technique that can be used to remove embedded watermarks from cover images. The study examines the EBLIND/DLC watermarking algorithm, which operates in the spatial domain and is designed to embed a 1-bit message into an image. To evaluate robustness against DCT-filtering, the research utilizes software implementations of the watermarking algorithms, the dctdnoiz filter from FFmpeg, and the Flickr8k image database, which contains 8091 images. An extension to the E_BLIND/D_LC algorithm is proposed to enhance its robustness against DCT-filtering. The paper presents an experiment comparing the robustness of the extended algorithm with the original E_BLIND/D_LC. The results demonstrate that the extended algorithm improves robustness against DCT-filtering and also maintains better image quality, as measured by the PSNR metric, at equal levels of robustness in terms of the number of successful watermark detections after filtering. This suggests that the proposed extension effectively balances enhanced robustness against DCT-filtering with the preservation of image quality. The experiment confirms that the extended version outperforms the original algorithm in both detection reliability and image quality under DCT-filtering conditions. The paper includes links to the algorithm implementations on C++ programming language and the Python testing scripts, which can be useful for further research and practical applications in software development.