Date: October 10, 2025

Author:  Huo Yanhao

A research paper titled “DRSW: Dual-stage Robust Semantic Watermarking for Image Semantic Communication” by Professor Xiang Shijun’s team from the College of Information Science and Technology has been officially accepted and published online in the internationally renowned journal IEEE Transactions on Circuits and Systems for Video Technology (IEEE TCSVT) (as shown in Figure 1). The work presents an innovative integration of information hiding and artificial intelligence technologies for image semantic communication.

The first author of the paper is Huo Yanhao, a Ph.D. student enrolled in 2023 in Cyberspace Security, and the corresponding author is Professor Xiang Shijun, a faculty member of the Department of Electronic Engineering and a doctoral supervisor in Cyberspace Security at the College of Information Science and Technology. IEEE TCSVT, established by IEEE in 1991, is a leading international journal in the field of video and image technology. It is recognized as a Chinese Academy of Sciences Zone 1 TOP journal with a latest impact factor of 11.1.

Semantic communication enhances the efficiency of information exchange by transmitting compact semantic representations, showing significant potential in applications such as autonomous driving and medical diagnosis. However, existing copyright protection methods face major limitations: traditional transform-domain watermarks often become ineffective during semantic extraction, while deep learning-based approaches lack robustness when integrated with semantic communication systems. More critically, current solutions fail to adequately protect the intellectual property of semantic information itself.

To address these challenges, the research team proposed a Dual-stage Robust Semantic Watermarking (DRSW) framework (as shown in Figure 2). By embedding watermarks in the frequency domain of semantic features, DRSW enables simultaneous copyright protection for both semantic content and reconstructed images. The framework maintains semantic consistency and high image quality while demonstrating strong robustness against channel noise, offering a new approach to copyright protection in future semantic communication systems.

This research was supported by the National Natural Science Foundation of China General Project “Research on New Reversible Data Hiding Methods Based on Deep Learning,” a sub-project of the National Key R&D Program of China “Digital Cultural Creation Asset Ownership Management and Service Network,” and the Guangdong Provincial Natural Science Foundation General Project “Research on New Robust Watermarking Technologies for Digital Image Copyright Protection.”

Link to the paper: https://ieeexplore.ieee.org/abstract/document/11178071

Editor: Zhou Huiqian