DRSW: Dual-stage Robust Semantic Watermarking for Image Semantic Communication

Semantic communication (SC) enables efficient information exchange by transmitting compact semantic representations rather than raw data, benefiting applications like autonomous driving and medical diagnosis. However, existing copyright protection methods face two key limitations: traditional transform-domain watermarking fails during semantic extraction, while deep learning-based methods lose robustness when integrated with SC. Most critically, existing solutions cannot protect semantic information itself, the core intellectual property in SC. To address these issues, we propose “Dual-stage Robust Semantic Watermarking” (DRSW), a framework that simultaneously protects the copyright for both semantics and reconstructed images. By embedding a watermark into the frequency domain of semantics, DRSW exhibits high robustness against possible channel noises while preserving semantic consistency and maintaining the reconstruction quality of images. Our work provides a new watermarking paradigm for future copyright protection in SC scenarios.