DefenseSplat: Enhancing the Robustness of 3D Gaussian Splatting via Frequency-Aware Filtering

2026-02-22

AI Summary

This work addresses the high sensitivity of 3D Gaussian Splatting (3DGS) to adversarial perturbations in input viewpoints, which can severely degrade rendering quality, drastically increase computational overhead, or even cause denial of service. For the first time, this study systematically reveals and analyzes the vulnerability of 3DGS under such attacks and proposes a frequency-aware defense mechanism that operates without access to clean ground-truth supervision. The method leverages wavelet transform to decouple image components into low- and high-frequency bands, preserving essential low-frequency structural information while filtering out high-frequency adversarial noise. Extensive evaluations demonstrate that the proposed approach substantially enhances the robustness of 3DGS across various attack strengths and benchmarks, while maintaining excellent reconstruction fidelity on clean inputs.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

3D Gaussian Splatting (3DGS) has emerged as a powerful paradigm for real-time and high-fidelity 3D reconstruction from posed images. However, recent studies reveal its vulnerability to adversarial corruptions in input views, where imperceptible yet consistent perturbations can drastically degrade rendering quality, increase training and rendering time, and inflate memory usage, even leading to server denial-of-service. In our work, to mitigate this issue, we begin by analyzing the distinct behaviors of adversarial perturbations in the low- and high-frequency components of input images using wavelet transforms. Based on this observation, we design a simple yet effective frequency-aware defense strategy that reconstructs training views by filtering high-frequency noise while preserving low-frequency content. This approach effectively suppresses adversarial artifacts while maintaining the authenticity of the original scene. Notably, it does not significantly impair training on clean data, achieving a desirable trade-off between robustness and performance on clean inputs. Through extensive experiments under a wide range of attack intensities on multiple benchmarks, we demonstrate that our method substantially enhances the robustness of 3DGS without access to clean ground-truth supervision. By highlighting and addressing the overlooked vulnerabilities of 3D Gaussian Splatting, our work paves the way for more robust and secure 3D reconstructions.

Problem

• 3D Gaussian Splatting
• adversarial corruptions
• robustness
• 3D reconstruction
• frequency-aware filtering

Innovation

• 3D Gaussian Splatting
• adversarial robustness
• frequency-aware filtering
• wavelet transform
• defense mechanism