D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation

The susceptibility of deep neural networks (DNNs) to adversarial examples has prompted an increase in the deployment of adversarial attacks. Image-agnostic universal adversarial perturbations (UAPs) are much more threatening, but many limitations exist to implementing UAPs in real-world scenarios wh...

Full description

Saved in:
Bibliographic Details
Main Authors: Geunhyeok Yu, Minwoo Jeon, Hyoseok Hwang
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10542123/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849421753661521920
author Geunhyeok Yu
Minwoo Jeon
Hyoseok Hwang
author_facet Geunhyeok Yu
Minwoo Jeon
Hyoseok Hwang
author_sort Geunhyeok Yu
collection DOAJ
description The susceptibility of deep neural networks (DNNs) to adversarial examples has prompted an increase in the deployment of adversarial attacks. Image-agnostic universal adversarial perturbations (UAPs) are much more threatening, but many limitations exist to implementing UAPs in real-world scenarios where only binary decisions are returned. In this research, we propose D-BADGE, a novel method to craft universal adversarial perturbations for executing decision- To primarily optimize perturbation by focusing on decisions, we consider the direction of these updates as the primary factor and the magnitude of updates as the secondary factor. First, we employ Hamming loss that measures the distance from distributions of ground truth and accumulating decisions in batches to determine the magnitude of the gradient. This magnitude is applied in the direction of the revised simultaneous perturbation stochastic approximation (SPSA) to update the perturbation. This simple yet efficient decision-based method functions similarly to a score-based attack, enabling the generation of UAPs in real-world scenarios, and can be easily extended to targeted attacks. Experimental validation across multiple victim models demonstrates that the D-BADGE outperforms existing attack methods, even image-specific and score-based attacks. In particular, our proposed method shows a superior attack success rate with less training time. The research also shows that D-BADGE can successfully deceive unseen victim models and accurately target specific classes.
format Article
id doaj-art-6d44444686fe49619c299c24d64d7b46
institution Kabale University
issn 2169-3536
language English
publishDate 2024-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-6d44444686fe49619c299c24d64d7b462025-08-20T03:31:23ZengIEEEIEEE Access2169-35362024-01-0112807708078010.1109/ACCESS.2024.340709710542123D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient EstimationGeunhyeok Yu0https://orcid.org/0009-0003-4912-2084Minwoo Jeon1Hyoseok Hwang2https://orcid.org/0000-0003-3241-8455Department of Software Convergence, Kyung Hee University, Yongin-si, Republic of KoreaDepartment of Software Convergence, Kyung Hee University, Yongin-si, Republic of KoreaDepartment of Software Convergence, Kyung Hee University, Yongin-si, Republic of KoreaThe susceptibility of deep neural networks (DNNs) to adversarial examples has prompted an increase in the deployment of adversarial attacks. Image-agnostic universal adversarial perturbations (UAPs) are much more threatening, but many limitations exist to implementing UAPs in real-world scenarios where only binary decisions are returned. In this research, we propose D-BADGE, a novel method to craft universal adversarial perturbations for executing decision- To primarily optimize perturbation by focusing on decisions, we consider the direction of these updates as the primary factor and the magnitude of updates as the secondary factor. First, we employ Hamming loss that measures the distance from distributions of ground truth and accumulating decisions in batches to determine the magnitude of the gradient. This magnitude is applied in the direction of the revised simultaneous perturbation stochastic approximation (SPSA) to update the perturbation. This simple yet efficient decision-based method functions similarly to a score-based attack, enabling the generation of UAPs in real-world scenarios, and can be easily extended to targeted attacks. Experimental validation across multiple victim models demonstrates that the D-BADGE outperforms existing attack methods, even image-specific and score-based attacks. In particular, our proposed method shows a superior attack success rate with less training time. The research also shows that D-BADGE can successfully deceive unseen victim models and accurately target specific classes.https://ieeexplore.ieee.org/document/10542123/Deep neural networksuniversal decision-based adversarial attackimage classificationrepresentation learningvulnerabilityzeroth-order optimization
spellingShingle Geunhyeok Yu
Minwoo Jeon
Hyoseok Hwang
D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation
IEEE Access
Deep neural networks
universal decision-based adversarial attack
image classification
representation learning
vulnerability
zeroth-order optimization
title D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation
title_full D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation
title_fullStr D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation
title_full_unstemmed D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation
title_short D-BADGE: Decision-Based Adversarial Batch Attack With Directional Gradient Estimation
title_sort d badge decision based adversarial batch attack with directional gradient estimation
topic Deep neural networks
universal decision-based adversarial attack
image classification
representation learning
vulnerability
zeroth-order optimization
url https://ieeexplore.ieee.org/document/10542123/
work_keys_str_mv AT geunhyeokyu dbadgedecisionbasedadversarialbatchattackwithdirectionalgradientestimation
AT minwoojeon dbadgedecisionbasedadversarialbatchattackwithdirectionalgradientestimation
AT hyoseokhwang dbadgedecisionbasedadversarialbatchattackwithdirectionalgradientestimation