Research map / Architectures and efficiency / RNNs and normalization
GlyphShield: Document Watermarking for the Physical World via Vector Typeface Synthesis
Nan Sun, Yuxing Lu, Han Fang, Hefei Ling, Sijing Xie, LuYu Yuan, Chengxin Zhao
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 |
|---|---|---|---|
| train font synthesis and decoder end-to-end with SDF differentiable rasterization (vs two-stage AutoStegaFont) | font watermarking · Song, Gaussian noise | bit accuracy (%): 60.57 -> 96.25 | Improved |
| GlyphShield vs AutoStegaFont under perspective transform | Times New Roman · Perspective (f=0.2) | bit accuracy (%): 98.07 -> 96.97 | Hurt |
| add Local branch to Global-only encoder (ablation) | Vector Encoder · Times New Roman | PSNR (dB): 30.42 -> 33.53 | Improved |
| add Global branch to Local-only encoder (ablation) | Vector Encoder · Times New Roman | bit accuracy (%): 82.73 -> 99.14 | Improved |
| add Local branch to Global-only encoder (ablation) | Vector Encoder · Times New Roman | bit accuracy (%): 96.07 -> 99.14 | Improved |
| larger font size in screenshots | GlyphShield · Song screenshot | bit accuracy (%): 63.76 -> 100 | Improved |
| GlyphShield vs CIN image watermarking under geometric transforms | character images · Rotation 10deg | bit accuracy (%): 57.86 -> 94.58 | Improved |
Closest papers by meaning
- TRACE: Structure-Aware Character Encoding for Robust and Generalizable Document Watermarking (NeurIPS 2026 · similarity 0.85)
- VecGlypher: Unified Vector Glyph Generation with Language Models (CVPR 2026 · similarity 0.84)
- Can Simple Averaging Defeat Modern Watermarks? (NeurIPS 2024 · similarity 0.83)
- ArcMark: Distortion-Free Multi-Byte LLM Watermark via Optimal Transport (NeurIPS 2026 · similarity 0.83)
- Revealing Weaknesses in Text Watermarking Through Self-Information Rewrite Attacks (ICML 2025 · similarity 0.83)
- Black-Box Detection of Language Model Watermarks (ICLR 2025 · similarity 0.83)
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.
