Reproduce “GTSD: Generative Text Steganography Based on Diffusion Model” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zhengxian Wu, Juan Wen, Yiming Xue, Ziwei Zhang, Yinghan Zhou
A new generative text steganography method, GTSD, leverages diffusion models to produce stego texts 50-100 times faster than prior approaches. It significantly enhances robustness against replacement attacks, achieving over 95% correct secret information extraction, and improves imperceptibility while maintaining strong anti-steganalysis capabilities.
Key insights
- The non-autoregressive nature of diffusion models fundamentally addresses the speed limitations of prior generative text steganography, enabling much faster parallel text generation.
- Combining batch generation with similarity-based extraction provides a robust mechanism for recovering hidden information, making stego texts resilient to minor modifications.
- The proposed prompt and batch mapping strategy allows for flexible and scalable embedding capacity without compromising imperceptibility or anti-steganalysis security.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.