Reversible Self-Embedding Watermarking for Encrypted Image Authentication

Image authentication is essential for detecting unauthorized modifications, yet privacy concerns often prevent image owners from disclosing sensitive visual content. This paper proposes a reversible self-embedding watermarking scheme for encrypted image authentication and tamper localization. The image owner encrypts the original image before transmission, enabling a third-party verifier to authenticate image integrity directly in the encrypted domain without accessing the original image or requiring an external reference. The proposed scheme detects tampering through watermark distortions and employs a hierarchical detection mechanism to reduce false negatives, increasing the theoretical detection rate from 87.5% to approximately 99.86%. The scheme achieves a pure embedding rate of 0.75 bits per pixel (bpp) and an average modification rate of 0.21875 modifications per pixel. Furthermore, the image owner can directly decrypt the watermarked encrypted image and recover the original image completely and losslessly without removing the watermark. Experimental results demonstrate the feasibility and efficiency of the proposed approach, offering privacy-preserving authentication, tamper localization, and direct lossless image recovery for secure image management.