Submit

Login Login

Register Register

Home / Journals / CMC / Online First / doi:10.32604/cmc.2026.079408
Journal Menu
Special Issues
Table of Content

All issues

Online First

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Open Access

ARTICLE

CycleGAN-RRW: Blind Reversible Image Watermarking via Cycle-Consistent Adversarial Feature Encoding for Secure Image Ownership Authentication

Mohammed Shamar Yadkar1, Sefer Kurnaz1, Saadaldeen Rashid Ahmed2,3,*
1 Department of Electrical and Computer Engineering, Altinbas University, Istanbul, Türkiye
2 Artificial Intelligence Engineering Department, College of Engineering, Al-Ayen University, Nasiriyah, Thi-Qar, Iraq
3 Computer Science, Bayan University, Erbil, Kurdistan, Iraq
* Corresponding Author: Saadaldeen Rashid Ahmed. Email: email
(This article belongs to the Special Issue: Development and Application of Deep Learning and Image Processing)

Computers, Materials & Continua https://doi.org/10.32604/cmc.2026.079408

Received 21 January 2026; Accepted 26 February 2026; Published online 20 March 2026

Abstract

This advanced research describes CycleGAN-RRW, a new reversible watermarking system for secure image ownership authentication. It uses Cycle-Consistent Generative Adversarial Networks with adaptive feature encoding. In areas such as law, forensics, and telemedicine, digital images usually contain private info that may be changed or used without authorization. Existing watermarking methods may decrease image quality, may not be reversible, or need outside keys. To address these problems, our model embeds metadata into intermediate feature maps with Adaptive Instance Normalization (AdaIN), based on adversarial and perceptual loss. The dual-generator design permits two-way translation between original and watermarked images, with pixel-level reversibility and semantic integrity. Key aims include blind watermark verification, eliminating side-channel dependency, and resisting distortions such as compression and noise. We tested our approach on the DIV2K and USC-SIPI Miscellaneous datasets, which showed acceptable watermark fidelity and reconstruction accuracy. The model achieved a Peak Signal-to-Noise Ratio (PSNR) of over 42 dB, a Structural Similarity Index (SSIM) above 0.98, and a Bit Error Rate (BER) below 1.5% when subjected to typical attacks like JPEG compression (Q ≥ 60) and Gaussian noise (σ = 5). The system permits watermark recovery and tamper detection without outside keys, with an ownership verification accuracy of 98.63%. The CycleGAN-RRW method is a self-contained, blind, and legally defensible watermarking solution with real-time inference and may be applied to other fields like forensic imaging and tele-health.

Keywords

Reversible watermarking; image ownership verification; deep learning; CycleGAN-RRW; image processing

  • 0

    View

  • 0

    Download

  • 0

    Like

Related articles
Copyright© 2026 Tech Science Press
© 1997-2026 TSP (Henderson, USA) unless otherwise stated

Share Link