StealthAttack: Robust 3D Gaussian Splatting Poisoning via Density-Guided Illusions
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908573730078720 |
|---|---|
| author | Ke, Bo-Hsu Xie, You-Zhe Liu, Yu-Lun Chiu, Wei-Chen |
| author_facet | Ke, Bo-Hsu Xie, You-Zhe Liu, Yu-Lun Chiu, Wei-Chen |
| contents | 3D scene representation methods like Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have significantly advanced novel view synthesis. As these methods become prevalent, addressing their vulnerabilities becomes critical. We analyze 3DGS robustness against image-level poisoning attacks and propose a novel density-guided poisoning method. Our method strategically injects Gaussian points into low-density regions identified via Kernel Density Estimation (KDE), embedding viewpoint-dependent illusory objects clearly visible from poisoned views while minimally affecting innocent views. Additionally, we introduce an adaptive noise strategy to disrupt multi-view consistency, further enhancing attack effectiveness. We propose a KDE-based evaluation protocol to assess attack difficulty systematically, enabling objective benchmarking for future research. Extensive experiments demonstrate our method's superior performance compared to state-of-the-art techniques. Project page: https://hentci.github.io/stealthattack/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_02314 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | StealthAttack: Robust 3D Gaussian Splatting Poisoning via Density-Guided Illusions Ke, Bo-Hsu Xie, You-Zhe Liu, Yu-Lun Chiu, Wei-Chen Computer Vision and Pattern Recognition 3D scene representation methods like Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) have significantly advanced novel view synthesis. As these methods become prevalent, addressing their vulnerabilities becomes critical. We analyze 3DGS robustness against image-level poisoning attacks and propose a novel density-guided poisoning method. Our method strategically injects Gaussian points into low-density regions identified via Kernel Density Estimation (KDE), embedding viewpoint-dependent illusory objects clearly visible from poisoned views while minimally affecting innocent views. Additionally, we introduce an adaptive noise strategy to disrupt multi-view consistency, further enhancing attack effectiveness. We propose a KDE-based evaluation protocol to assess attack difficulty systematically, enabling objective benchmarking for future research. Extensive experiments demonstrate our method's superior performance compared to state-of-the-art techniques. Project page: https://hentci.github.io/stealthattack/ |
| title | StealthAttack: Robust 3D Gaussian Splatting Poisoning via Density-Guided Illusions |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2510.02314 |