RenderBender: A Survey on Adversarial Attacks Using Differentiable Rendering

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hull, Matthew, Wang, Haoran, Lau, Matthew, Helbling, Alec, Phute, Mansi, Zhang, Chao, Kira, Zsolt, Lunardi, Willian, Andreoni, Martin, Lee, Wenke, Chau, Polo
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909629930274816
author Hull, Matthew
Wang, Haoran
Lau, Matthew
Helbling, Alec
Phute, Mansi
Zhang, Chao
Kira, Zsolt
Lunardi, Willian
Andreoni, Martin
Lee, Wenke
Chau, Polo
author_facet Hull, Matthew
Wang, Haoran
Lau, Matthew
Helbling, Alec
Phute, Mansi
Zhang, Chao
Kira, Zsolt
Lunardi, Willian
Andreoni, Martin
Lee, Wenke
Chau, Polo
contents Differentiable rendering techniques like Gaussian Splatting and Neural Radiance Fields have become powerful tools for generating high-fidelity models of 3D objects and scenes. Their ability to produce both physically plausible and differentiable models of scenes are key ingredient needed to produce physically plausible adversarial attacks on DNNs. However, the adversarial machine learning community has yet to fully explore these capabilities, partly due to differing attack goals (e.g., misclassification, misdetection) and a wide range of possible scene manipulations used to achieve them (e.g., alter texture, mesh). This survey contributes the first framework that unifies diverse goals and tasks, facilitating easy comparison of existing work, identifying research gaps, and highlighting future directions - ranging from expanding attack goals and tasks to account for new modalities, state-of-the-art models, tools, and pipelines, to underscoring the importance of studying real-world threats in complex scenes.
format Preprint
id arxiv_https___arxiv_org_abs_2411_09749
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle RenderBender: A Survey on Adversarial Attacks Using Differentiable Rendering
Hull, Matthew
Wang, Haoran
Lau, Matthew
Helbling, Alec
Phute, Mansi
Zhang, Chao
Kira, Zsolt
Lunardi, Willian
Andreoni, Martin
Lee, Wenke
Chau, Polo
Machine Learning
Cryptography and Security
Computer Vision and Pattern Recognition
Differentiable rendering techniques like Gaussian Splatting and Neural Radiance Fields have become powerful tools for generating high-fidelity models of 3D objects and scenes. Their ability to produce both physically plausible and differentiable models of scenes are key ingredient needed to produce physically plausible adversarial attacks on DNNs. However, the adversarial machine learning community has yet to fully explore these capabilities, partly due to differing attack goals (e.g., misclassification, misdetection) and a wide range of possible scene manipulations used to achieve them (e.g., alter texture, mesh). This survey contributes the first framework that unifies diverse goals and tasks, facilitating easy comparison of existing work, identifying research gaps, and highlighting future directions - ranging from expanding attack goals and tasks to account for new modalities, state-of-the-art models, tools, and pipelines, to underscoring the importance of studying real-world threats in complex scenes.
title RenderBender: A Survey on Adversarial Attacks Using Differentiable Rendering
topic Machine Learning
Cryptography and Security
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.09749