MultiNeRF: Multiple Watermark Embedding for Neural Radiance Fields

2025-04-03

AI Summary

Existing NeRF watermarking methods suffer from coupling between multiple embedded watermarks and scene representation, degrading rendering fidelity. Method: We propose the first differentiable watermarking framework supporting concurrent multi-key watermarking. Specifically, we design a dedicated watermark voxel grid (extended from TensoRF) and introduce a FiLM-based conditional modulation mechanism to fully decouple watermark signals from scene representation. An end-to-end differentiable embedding/extraction pipeline enables dynamic activation and extraction of any watermark by identifier—without retraining. Results: On NeRF-Synthetic and LLFF datasets, our method achieves statistically insignificant degradation in PSNR and SSIM (p > 0.05), while significantly increasing watermark capacity (p < 0.01). It supports concurrent embedding and high-fidelity extraction of ≥8 independent key-based watermarks, jointly ensuring high-rendering fidelity and flexible copyright management.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• Responsible Web: Algorithmic accountability and transparency on the web

Abstract

We present MultiNeRF, a 3D watermarking method that embeds multiple uniquely keyed watermarks within images rendered by a single Neural Radiance Field (NeRF) model, whilst maintaining high visual quality. Our approach extends the TensoRF NeRF model by incorporating a dedicated watermark grid alongside the existing geometry and appearance grids. This extension ensures higher watermark capacity without entangling watermark signals with scene content. We propose a FiLM-based conditional modulation mechanism that dynamically activates watermarks based on input identifiers, allowing multiple independent watermarks to be embedded and extracted without requiring model retraining. MultiNeRF is validated on the NeRF-Synthetic and LLFF datasets, with statistically significant improvements in robust capacity without compromising rendering quality. By generalizing single-watermark NeRF methods into a flexible multi-watermarking framework, MultiNeRF provides a scalable solution for 3D content. attribution.

Problem

• Embed multiple watermarks in NeRF without quality loss
• Separate watermark signals from 3D scene content
• Enable dynamic watermark activation via input identifiers

Innovation

• Extends TensoRF with dedicated watermark grid
• Uses FiLM-based conditional modulation mechanism
• Enables multiple watermark embedding without retraining