Research map / Video / Video generation
SPDMark: Selective Parameter Displacement for Robust Video Watermarking
Samar Fares, Nurbek Tastan, Karthik Nandakumar
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 |
|---|---|---|---|
| increase generated frames from 8 to 32 (ablation) | SPDMark · Robustness on SVD-XT | Robust Acc: 0.880 -> 0.933 | Improved |
| increase generated frames from 8 to 32 (ablation) | SPDMark · Clean extraction on SVD-XT | Bit Acc: 0.942 -> 0.996 | Improved |
| increase generated frames from 8 to 32 (ablation) | SPDMark · Video quality on SVD-XT | Imaging Quality: 0.721 -> 0.700 | Hurt |
| raise diffusion steps from 10 to 50 (ablation) | SPDMark · Clean extraction on SVD-XT | Bit Acc: 0.989 -> 0.995 | No clear effect |
| raise diffusion steps from 10 to 50 (ablation) | SPDMark · Video quality on SVD-XT | Imaging Quality: 0.630 -> 0.713 | Improved |
| raise guidance scale from 1 to 5 (ablation) | SPDMark · Clean extraction on SVD-XT | Bit Acc: 0.998 -> 0.991 | No clear effect |
| increase payload from 28 to 56 bits per frame under fixed training budget (ablation) | SPDMark · Extraction accuracy on SVD-XT | Bit Acc and Robust Acc | Hurt |
| train higher-payload variant 12k extra steps (G3) (ablation) | SPDMark · Extraction on SVD-XT | Clean/Robust Acc: clean 0.969, robust 0.890 | Improved |
| use SPDMark instead of baselines for in-generation watermarking | Video diffusion models · Robustness average on SVD-XT | Bit Acc: 0.912 -> 0.935 | Improved |
| use SPDMark instead of VideoSeal for clean extraction | Video diffusion models · Clean extraction on SVD-XT | Bit Acc: 0.999 -> 0.995 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- VideoSeal — VideoSEAL: Mitigating Evidence Misalignment in Agentic Long Video Understanding by Decoupling Answer Authority (ICML 2026) · better on 4 of 6 shared results
- VideoShield — VideoShield: Regulating Diffusion-based Video Generation Models via Watermarking (ICLR 2025) · better on 5 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- VideoShield: Regulating Diffusion-based Video Generation Models via Watermarking (ICLR 2025 · similarity 0.89)
- Safe-Sora: Safe Text-to-Video Generation via Graphical Watermarking (NeurIPS 2025 · similarity 0.89)
- BitMark: Watermarking Bitwise Autoregressive Image Generative Models (NeurIPS 2025 · similarity 0.86)
- T2SMark: Balancing Robustness and Diversity in Noise-as-Watermark for Diffusion Models (NeurIPS 2025 · similarity 0.86)
- A Locally Tokenized Generative Model for Robust Time-Series Watermarking (NeurIPS 2026 · similarity 0.86)
- Robust Watermarking Using Generative Priors Against Image Editing: From Benchmarking to Advances (ICLR 2025 · similarity 0.85)
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.
