Research map / Image generation / Autoregressive and latent generation
GVIS: Generative Vector Image Steganography
Zihao Xu, Dawei Xu, Zihan Li, Xixi Zheng, Chuan Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use GVIS generative vectorization instead of prior vector steganography | vector image steganography · CelebA-HQ 256x256 | PSNR: 63.87 -> 67.87 (vs StegoBIT) | Improved |
| use GVIS instead of prior vector methods | vector image steganography · CelebA-HQ 256x256 | capacity (bits): 48000 -> 88878 (vs StegoSVG) | Improved |
| use GVIS instead of prior vector methods | vector image steganography · LSUN-Bedrooms 256x256 | PSNR: 63.10 -> 69.50 (vs StegoBIT) | Improved |
| use GVIS instead of StegoSVG and StegoBIT | SVG steganography · CelebA-HQ average SVG file size | file size (KB): 1485.24 KB (StegoSVG) -> 519.99 KB | Improved |
| increase angle partition from 2^20 to 2^24 (length 24) (ablation) | GVIS · CelebA-HQ capacity vs accuracy | bit accuracy: 1.0000 -> 0.9999 | No clear effect |
| increase angle partition to 2^28 (more capacity) (ablation) | GVIS · CelebA-HQ capacity vs accuracy | bit accuracy: 1.0000 -> 0.9750 | Hurt |
| increase partition granularity from 2^4 to 2^20 angle bins (ablation) | GVIS · CelebA-HQ capacity vs accuracy | capacity (bits): 18516 -> 88878 | Improved |
| apply steganalysis networks to rasterized GVIS stego images | GVIS · CelebA-HQ steganalysis (SiaStegNet, StegNet, ZhuNet, YeNet, XuNet) | detection accuracy: around 50% | No clear effect |
| add overlap detection to exclude overlapping curves from embedding | GVIS · security analysis | steganalysis risk | Improved |
Closest papers by meaning
- MDDM: Practical Message-Driven Generative Image Steganography Based on Diffusion Models (ICML 2025 · similarity 0.88)
- Neural Cover Selection for Image Steganography (NeurIPS 2024 · similarity 0.85)
- StegaStyleGAN: Towards Generic and Practical Generative Image Steganography (AAAI 2024 · similarity 0.85)
- Hiding Images in Diffusion Models by Editing Learned Score Functions (CVPR 2025 · similarity 0.84)
- StegoGAN: Leveraging Steganography for Non-Bijective Image-to-Image Translation (CVPR 2024 · similarity 0.84)
- SSHR: More Secure Generative Steganography with High-Quality Revealed Secret Images (ICML 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
