GaussianSeal: Rooting Adaptive Watermarks for 3D Gaussian Generation Model

2025-03-01

AI Summary

3D Gaussian Splatting (3DGS) generative models lack effective copyright protection mechanisms. Method: This paper introduces the first end-to-end trainable bit-level watermarking framework tailored for 3D generative content. It proposes an adaptive bit modulation module to jointly optimize watermark embedding and 3DGS generation, employs network-block-level embedding, and incorporates a 3DGS rendering consistency constraint to preserve geometric and visual fidelity. Contribution/Results: The framework achieves near-lossless generation quality (PSNR > 35 dB) while attaining >99.2% bit decoding accuracy—significantly outperforming conventional post-hoc watermarking approaches. This work bridges a critical research gap in copyright protection for 3D generative models and establishes a novel paradigm for intellectual property attribution of 3D content in the AIGC era.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

With the advancement of AIGC technologies, the modalities generated by models have expanded from images and videos to 3D objects, leading to an increasing number of works focused on 3D Gaussian Splatting (3DGS) generative models. Existing research on copyright protection for generative models has primarily concentrated on watermarking in image and text modalities, with little exploration into the copyright protection of 3D object generative models. In this paper, we propose the first bit watermarking framework for 3DGS generative models, named GaussianSeal, to enable the decoding of bits as copyright identifiers from the rendered outputs of generated 3DGS. By incorporating adaptive bit modulation modules into the generative model and embedding them into the network blocks in an adaptive way, we achieve high-precision bit decoding with minimal training overhead while maintaining the fidelity of the model's outputs. Experiments demonstrate that our method outperforms post-processing watermarking approaches for 3DGS objects, achieving superior performance of watermark decoding accuracy and preserving the quality of the generated results.

Problem

• Develops a watermarking framework for 3D Gaussian Splatting models.
• Addresses copyright protection for 3D object generative models.
• Ensures high-precision bit decoding with minimal training overhead.

Innovation

• First bit watermarking framework for 3DGS models
• Adaptive bit modulation in generative model blocks
• High-precision decoding with minimal training overhead