$S^2$NeRF: Privacy-preserving Training Framework for NeRF

TL;DR

This paper introduces $S^2$NeRF, a privacy-preserving training framework for Neural Radiance Fields that combines split learning with noise-based defenses to protect sensitive scene data during collaborative model training.

Contribution

The paper proposes $S^2$NeRF, a novel privacy-preserving NeRF training method that defends against reconstruction attacks using gradient noise, advancing secure 3D scene modeling.

Findings

$S^2$NeRF effectively defends against reconstruction attacks.

The framework maintains high model utility with privacy guarantees.

Extensive evaluations validate its security and performance.

Abstract

Neural Radiance Fields (NeRF) have revolutionized 3D computer vision and graphics, facilitating novel view synthesis and influencing sectors like extended reality and e-commerce. However, NeRF's dependence on extensive data collection, including sensitive scene image data, introduces significant privacy risks when users upload this data for model training. To address this concern, we first propose SplitNeRF, a training framework that incorporates split learning (SL) techniques to enable privacy-preserving collaborative model training between clients and servers without sharing local data. Despite its benefits, we identify vulnerabilities in SplitNeRF by developing two attack methods, Surrogate Model Attack and Scene-aided Surrogate Model Attack, which exploit the shared gradient data and a few leaked scene images to reconstruct private scene information. To counter these threats, we introduce $S^2$NeRF, secure SplitNeRF that integrates effective defense mechanisms. By introducing decaying noise related to the gradient norm into the shared gradient information, $S^2$NeRF preserves privacy while maintaining a high utility of the NeRF model. Our extensive evaluations across multiple datasets demonstrate the effectiveness of $S^2$NeRF against privacy breaches, confirming its viability for secure NeRF training in sensitive applications.