Can Protective Watermarking Safeguard the Copyright of 3D Gaussian Splatting?

TL;DR

This work identifies vulnerabilities in existing 3D Gaussian Splatting watermarking schemes by showing that traditional 2D purification methods do not transfer to 3DGS. It introduces GSPure, a 3D-aware purification framework that uses view-dependent Gaussian contribution weights and geometry-based clustering to separate watermark-related Gaussians from the original scene, followed by adaptive pruning. Experiments on multiple 3DGS watermarking methods across the Mip-NeRF360 dataset demonstrate that GSPure achieves superior watermark purification with minimal degradation to scene fidelity, outperforming several baselines in both effectiveness and generalization. The results underscore the need for robust 3DGS watermarking and adaptive defenses against watermark purification attacks, shaping future work in copyright protection for 3D scene representations.

Abstract

3D Gaussian Splatting (3DGS) has emerged as a powerful representation for 3D scenes, widely adopted due to its exceptional efficiency and high-fidelity visual quality. Given the significant value of 3DGS assets, recent works have introduced specialized watermarking schemes to ensure copyright protection and ownership verification. However, can existing 3D Gaussian watermarking approaches genuinely guarantee robust protection of the 3D assets? In this paper, for the first time, we systematically explore and validate possible vulnerabilities of 3DGS watermarking frameworks. We demonstrate that conventional watermark removal techniques designed for 2D images do not effectively generalize to the 3DGS scenario due to the specialized rendering pipeline and unique attributes of each gaussian primitives. Motivated by this insight, we propose GSPure, the first watermark purification framework specifically for 3DGS watermarking representations. By analyzing view-dependent rendering contributions and exploiting geometrically accurate feature clustering, GSPure precisely isolates and effectively removes watermark-related Gaussian primitives while preserving scene integrity. Extensive experiments demonstrate that our GSPure achieves the best watermark purification performance, reducing watermark PSNR by up to 16.34dB while minimizing degradation to original scene fidelity with less than 1dB PSNR loss. Moreover, it consistently outperforms existing methods in both effectiveness and generalization.

Figures (5)

• Figure 1: Visualization of purification results by GSPure on existing scene-level watermarking techniques guo2024splatszhang2024gshider. From left to right: watermark extracted by the copyright owner from the original published point cloud; the original published point cloud containing watermark information; feature clustering results which highlights watermark-correlated Gaussian primitives achieved by our GSPure (lower ranks indicate lower preservation priority); and the watermark extracted after applying our adaptive pruning. Clearly, the majority of watermark information is effectively removed.
• Figure 2: Overview of our GSPure framework.First, we compute the rendering contribution of each Gaussian primitive from multiple viewpoints. Next, we construct geometry-aware features and use HDBSCAN to identify watermark-correlated clusters. Finally, adaptive pruning based on cluster-level and noise-level thresholds ($\tau_c$ and $\tau_n$) effectively removes watermark-related Gaussians while preserving original scene fidelity.
• Figure 3: Qualitative comparisons on Mipnerf360 datasets. The first row of each group represents the original scene while the second represents the hidden watermark. Our GSPure effectively removes watermarks across all 3DGS watermarking methods while preserving the integrity of the original scene with minimal impact.
• Figure 4: Impact of noise and cluster pruning factors on scene quality and watermark removal.
• Figure 5: Visualization of Point Cloud. The left column represents the "Ground Truth". The middle column shows the clustering results of our GSPure. The right column presents the pruned point cloud after GSPure.