StealthAttack：通过密度引导幻象实现鲁棒的3D高斯泼溅投毒
StealthAttack: Robust 3D Gaussian Splatting Poisoning via Density-Guided Illusions
AI总结：
针对3D高斯泼溅（3DGS）的图像级投毒脆弱性问题，提出基于核密度估计（KDE）的密度引导投毒方法与自适应噪声策略，可实现隐蔽且有效的3DGS投毒攻击，并建立客观评估基准。
摘要：
神经辐射场（NeRF）、3D高斯泼溅（3DGS）等3D场景表示方法极大推动了新视角合成的发展。随着这类方法日益普及，应对其脆弱性变得至关重要。我们分析了3DGS针对图像级投毒攻击的鲁棒性，提出一种全新的密度引导投毒方法。该方法通过核密度估计（KDE）识别低密度区域，策略性地将高斯点注入其中，嵌入依赖视角的幻象物体，使其在受投毒视角下清晰可见，同时对正常视角的影响极小。此外，我们提出一种自适应噪声策略破坏多视图一致性，进一步提升攻击效果。我们还提出基于KDE的评估协议，可系统评估攻击难度，为后续研究提供客观基准。大量实验表明，该方法性能优于现有最优技术。项目页面：https://hentci.github.io/stealthattack/
英文摘要：
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/
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
