Saved in:
Bibliographic Details
Main Authors: Jacob, Sven, Shao, Weijia, Kasneci, Gjergji
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.14460
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909849811419136
author Jacob, Sven
Shao, Weijia
Kasneci, Gjergji
author_facet Jacob, Sven
Shao, Weijia
Kasneci, Gjergji
contents Video-based object detection plays a vital role in safety-critical applications. While deep learning-based object detectors have achieved impressive performance, they remain vulnerable to adversarial attacks, particularly those involving universal perturbations. In this work, we propose a minimally distorted universal adversarial attack tailored for video object detection, which leverages nuclear norm regularization to promote structured perturbations concentrated in the background. To optimize this formulation efficiently, we employ an adaptive, optimistic exponentiated gradient method that enhances both scalability and convergence. Our results demonstrate that the proposed attack outperforms both low-rank projected gradient descent and Frank-Wolfe based attacks in effectiveness while maintaining high stealthiness. All code and data are publicly available at https://github.com/jsve96/AO-Exp-Attack.
format Preprint
id arxiv_https___arxiv_org_abs_2510_14460
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Structured Universal Adversarial Attacks on Object Detection for Video Sequences
Jacob, Sven
Shao, Weijia
Kasneci, Gjergji
Computer Vision and Pattern Recognition
Video-based object detection plays a vital role in safety-critical applications. While deep learning-based object detectors have achieved impressive performance, they remain vulnerable to adversarial attacks, particularly those involving universal perturbations. In this work, we propose a minimally distorted universal adversarial attack tailored for video object detection, which leverages nuclear norm regularization to promote structured perturbations concentrated in the background. To optimize this formulation efficiently, we employ an adaptive, optimistic exponentiated gradient method that enhances both scalability and convergence. Our results demonstrate that the proposed attack outperforms both low-rank projected gradient descent and Frank-Wolfe based attacks in effectiveness while maintaining high stealthiness. All code and data are publicly available at https://github.com/jsve96/AO-Exp-Attack.
title Structured Universal Adversarial Attacks on Object Detection for Video Sequences
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2510.14460