Semi-fragile image watermarking: A comprehensive survey, classification, and research directions

Computer Science Review. 2027. Vol. 63 (A). Article 101066.

Melman A., Evsyutin O.

One of the effective means of protecting digital images is invisible watermarking. Different scenarios have different requirements for watermark resistance to post-processing. Robust watermarking schemes are resistant to a wide range of modifications. Fragile watermarking schemes are designed for integrity control and are destroyed by even minor modifications of images. Semi-fragile watermarking schemes combine features of robust and fragile ones to solve image authentication problems. The design of such schemes ensures that the authenticity of images is confirmed under "harmless" modifications such as compression, brightness changes, blurring, or rotation. However, if the image content is tampered, for example, by inserting or deleting an object, image authenticity verification reveals the forgery. To date, researchers have proposed many different approaches to ensuring semi-fragility of watermark embedding for digital images, including the use of traditional embedding methods and the use of neural networks. In this study, we propose a classification of semi-fragile image watermarking schemes and a review of current studies in this field. We analyze known schemes in terms of watermark type, tamper detection approach, and the ability to localize and recover the tampered regions. We also discuss the effectiveness of the considered schemes in terms of embedding imperceptibility, robustness, and fragility. Based on the review results, we highlight the challenges in semi-fragile image watermarking and formulate promising research directions that may be useful to researchers in the field of developing digital image authentication tools.

Keywords: deepfake detection, image watermarking, semi-fragile watermarking, image authentication, tamper localization, tamper recovery