Research map / 3D vision / Gaussian splatting
StealthAttack: Robust 3D Gaussian Splatting Poisoning via Density-Guided Illusions
Bo-Hsu Ke, You-Zhe Xie, Yu-Lun Liu, Wei-Chen Chiu
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 density-guided point cloud attack to GT replacement (ablation) | 3DGS poisoning · Mip-NeRF 360 V-illusory | PSNR: 13.22 -> 26.01 | Improved |
| add density-guided point cloud attack to GT replacement (ablation) | 3DGS poisoning · Mip-NeRF 360 | ASR (PSNR): 0/7 -> 6/7 | Improved |
| add view consistency disruption to GT replacement + density-guided (ablation) | 3DGS poisoning · Mip-NeRF 360 V-illusory | PSNR: 26.01 -> 27.04 | Improved |
| add view consistency disruption to GT replacement + density-guided (ablation) | 3DGS poisoning · Mip-NeRF 360 | ASR (PSNR): 6/7 -> 7/7 | Improved |
| add view consistency disruption noise (ablation) | 3DGS poisoning · Mip-NeRF 360 V-test | PSNR: 29.40 -> 27.76 | Hurt |
| add view consistency disruption without density-guided points (ablation) | 3DGS poisoning · Mip-NeRF 360 V-illusory | PSNR: 13.22 -> 13.31 | No clear effect |
| raise initial noise sigma0 from 30 to 100 (linear decay) (ablation) | View consistency disruption · Mip-NeRF 360 V-illusory | PSNR: 26.47 -> 27.04 | Improved |
| raise initial noise sigma0 from 30 to 100 (linear decay) (ablation) | View consistency disruption · Mip-NeRF 360 V-test | PSNR: 28.43 -> 27.76 | Hurt |
| use cosine decay instead of linear at sigma0=100 (ablation) | View consistency disruption · Mip-NeRF 360 V-test | PSNR: 27.76 -> 26.96 | Hurt |
| set KDE bandwidth h=7.5 instead of h=10.0 (ablation) | Density-guided attack · Mip-NeRF 360 V-illusory | PSNR: 26.89 -> 27.04 | Improved |
| set KDE bandwidth h=7.5 instead of h=0.1 (ablation) | Density-guided attack · Mip-NeRF 360 V-illusory | PSNR: 27.00 -> 27.04 | No clear effect |
| use density-guided attack instead of IPA-NeRF (Instant-NGP) | 3DGS poisoning · Mip-NeRF 360 V-illusory | PSNR: 17.60 -> 27.04 | Improved |
Closest papers by meaning
- Poison-splat: Computation Cost Attack on 3D Gaussian Splatting (ICLR 2025 · similarity 0.88)
- SplatFormer: Point Transformer for Robust 3D Gaussian Splatting (ICLR 2025 · similarity 0.86)
- Quantifying and Alleviating Co-Adaptation in Sparse-View 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.86)
- SecureGS: Boosting the Security and Fidelity of 3D Gaussian Splatting Steganography (ICLR 2025 · similarity 0.86)
- CoCoGaussian: Leveraging Circle of Confusion for Gaussian Splatting from Defocused Images (CVPR 2025 · similarity 0.86)
- FewViewGS: Gaussian Splatting with Few View Matching and Multi-stage Training (NeurIPS 2024 · similarity 0.86)
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.
