Research map / Adversarial robustness and security / Attacks on text and graphs
SpecGuard: Spectral Projection-based Advanced Invisible Watermarking
Inzamamul Alam, Md Tanvir Islam, Simon S. Woo, Khan Muhammad
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 |
|---|---|---|---|
| add spectral projection (FFT approximation) on top of wavelet projection L1 (ablation) | SpecGuard · Ablation, 128-bit, no attack | BRA: 0.92 -> 0.99 | Improved |
| add spectral projection on top of wavelet projection L1 (ablation) | SpecGuard · Ablation, 128-bit, no attack | PSNR: 40.51 -> 42.89 | Improved |
| use wavelet level L1 instead of L2 with spectral projection (ablation) | SpecGuard · Ablation, 128-bit | PSNR: 36.25 -> 42.89 | Improved |
| add learnable threshold theta (Parseval-guided) to WP+SP (ablation) | SpecGuard · Flip attack | BRA: 0.48 -> 0.60 | Improved |
| add learnable threshold theta to WP+SP (ablation) | SpecGuard · Saturation +40% | BRA: 0.83 -> 0.92 | Improved |
| add learnable threshold theta to WP+SP (ablation) | SpecGuard · No attack | PSNR: 38.8 -> 42.9 | Improved |
| add spectral projection to wavelet projection (fixed theta) (ablation) | SpecGuard · Scale 0.75x | BRA: 0.65 -> 0.85 | Improved |
| use LeakyReLU instead of Tanh (ablation) | SpecGuard · Activation ablation, r=100 | BRA: 0.80 -> 0.99 | Improved |
| use LeakyReLU instead of ReLU (ablation) | SpecGuard · Activation ablation, r=100 | PSNR: 37.96 -> 42.89 | Improved |
| use radius 100 instead of 50 with LeakyReLU (ablation) | SpecGuard · Radius ablation | PSNR: 39.77 -> 42.89 | Improved |
| strength factor 20 instead of 5 (ablation) | SpecGuard · Strength factor ablation | PSNR: 40.79 -> 42.89 | Improved |
| strength factor 20 instead of 15 (ablation) | SpecGuard · Strength factor ablation | PSNR: 38.14 -> 42.89 | Improved |
Closest papers by meaning
- Hidden in the Noise: Two-Stage Robust Watermarking for Images (ICLR 2025 · similarity 0.85)
- Invisible Image Watermarks Are Provably Removable Using Generative AI (NeurIPS 2024 · similarity 0.84)
- WMCopier: Forging Invisible Watermarks on Arbitrary Images (NeurIPS 2025 · similarity 0.84)
- WateRF: Robust Watermarks in Radiance Fields for Protection of Copyrights (CVPR 2024 · similarity 0.84)
- Lightweight-Mark: Rethinking Deep Learning-Based Watermarking (ICML 2025 · similarity 0.84)
- Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models (NeurIPS 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.
