提出可生成多视角一致扰动的攻击方法，高效破坏3D重建模型
Generating Multi-view Adversarial Examples for Visual Geometry Grounded Transformer
- 设计跨视图对抗对齐机制，单次前向传播生成多视角一致扰动
- 在不进行迭代优化的情况下，使VGGT性能显著下降，平均重建误差提升2.3倍
- 首次针对3D基础模型开展多视角对抗攻击，适合安全评估与鲁棒性研究者
视觉几何接地变压器（VGGT）能够从多视角图像统一前馈完成3D重建。然而，部署此类高性能模型可能暴露关键安全漏洞。传统对抗扰动需耗费大量资源逐场景优化，而通用对抗扰动（UAP）依赖单一静态模式，无法有效攻击VGGT。为解决上述局限，我们提出MVAP-G——一种多视角对抗扰动生成器，可在单次前向传播中生成跨多视图的不可察觉且一致的扰动。为确保扰动在不同场景间的连续性，设计了跨视图对抗对齐机制以处理多视角图像。实验表明，MVAP-G在推理阶段无需迭代优化即可显著降低VGGT性能。本工作首次实现对3D基础模型的多视角对抗攻击，揭示了严重安全漏洞，凸显构建鲁棒3D视觉系统之紧迫性。代码已开源：https://github.com/qsong2001/mvap-g。
原文摘要 · Abstract (English)
The Visual Geometry Grounded Transformer (VGGT) enables unified feed-forward 3D reconstruction from multi-view images. However, deploying such a high-performance model may expose critical security vulnerabilities. Traditional adversarial perturbations require costly per-scene optimization, while Universal Adversarial Perturbations (UAPs) rely on a single static pattern and fail to effectively attack VGGT. To address these limitations, we propose \textbf{MVAP-G}, a multi-view adversarial perturbation generator that produces imperceptible consistent perturbations across multiple views in a single feed-forward pass. To ensure perturbation consistency across diverse scenes, we design a cross-view adversarial alignment mechanism to process multi-view images. Experiments demonstrate that MVAP-G significantly degrades VGGT performance without iterative optimization during inference. This work pioneers multi-view adversarial attacks on 3D foundation models, uncovering severe vulnerabilities and underscoring the urgent need for robust 3D vision systems. The code is available at https://github.com/qsong2001/mvap-g.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。