3D Gaussian Splat Vulnerabilities
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866912405544501248 |
|---|---|
| author | Hull, Matthew Yang, Haoyang Mehta, Pratham Phute, Mansi Cho, Aeree Wang, Haoran Lau, Matthew Lee, Wenke Lunardi, Willian T. Andreoni, Martin Chau, Polo |
| author_facet | Hull, Matthew Yang, Haoyang Mehta, Pratham Phute, Mansi Cho, Aeree Wang, Haoran Lau, Matthew Lee, Wenke Lunardi, Willian T. Andreoni, Martin Chau, Polo |
| contents | With 3D Gaussian Splatting (3DGS) being increasingly used in safety-critical applications, how can an adversary manipulate the scene to cause harm? We introduce CLOAK, the first attack that leverages view-dependent Gaussian appearances - colors and textures that change with viewing angle - to embed adversarial content visible only from specific viewpoints. We further demonstrate DAGGER, a targeted adversarial attack directly perturbing 3D Gaussians without access to underlying training data, deceiving multi-stage object detectors e.g., Faster R-CNN, through established methods such as projected gradient descent. These attacks highlight underexplored vulnerabilities in 3DGS, introducing a new potential threat to robotic learning for autonomous navigation and other safety-critical 3DGS applications. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_00280 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | 3D Gaussian Splat Vulnerabilities Hull, Matthew Yang, Haoyang Mehta, Pratham Phute, Mansi Cho, Aeree Wang, Haoran Lau, Matthew Lee, Wenke Lunardi, Willian T. Andreoni, Martin Chau, Polo Cryptography and Security Computer Vision and Pattern Recognition Machine Learning With 3D Gaussian Splatting (3DGS) being increasingly used in safety-critical applications, how can an adversary manipulate the scene to cause harm? We introduce CLOAK, the first attack that leverages view-dependent Gaussian appearances - colors and textures that change with viewing angle - to embed adversarial content visible only from specific viewpoints. We further demonstrate DAGGER, a targeted adversarial attack directly perturbing 3D Gaussians without access to underlying training data, deceiving multi-stage object detectors e.g., Faster R-CNN, through established methods such as projected gradient descent. These attacks highlight underexplored vulnerabilities in 3DGS, introducing a new potential threat to robotic learning for autonomous navigation and other safety-critical 3DGS applications. |
| title | 3D Gaussian Splat Vulnerabilities |
| topic | Cryptography and Security Computer Vision and Pattern Recognition Machine Learning |
| url | https://arxiv.org/abs/2506.00280 |