Query-Efficient Hard-Label Attack: A Prior-Guided Adam Ray Search Optimization
Sensors · 21.05.2026 · 26 (10) · 3272
DOI: 10.3390/s26103272 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 21.05.2026. Verfasst von Tianyi Ding und 4 weiteren Autor:innen. Beteiligte Einrichtung: Institute of Cyberspace Security, Zhejiang University of Technology, Hangzhou 310023, China. Die Arbeit nennt 25 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Deep neural networks are vulnerable to adversarial examples, even in hard-label black-box settings where only the top-1 prediction is available. To address the challenges of high-dimensional optimization under limited query budgets, we propose two query-efficient attack methods: Adam-OPT, which integrates Adam-based adaptive optimization into the ray-search framework to stabilize and accelerate zeroth-order gradient updates; Prior-Adam-OPT, which further incorporates transfer-based priors from surrogate models to enhance gradient estimation. Adam-OPT leverages historical gradient information and per-parameter adaptive updates to improve convergence, while Prior-Adam-OPT constructs a prior-guided orthogonal search basis that combines surrogate and random directions, enhancing both gradient accuracy and query efficiency. Our approach demonstrates superior performance across CIFAR-10, ImageNet, and zero-shot CLIP models, consistently reducing perturbation magnitudes and improving attack efficiency compared to state-of-the-art hard-label attacks. Ablation studies highlight the importance of the number of vectors used for gradient estimation and the quality of surrogate models, showing that combining adaptive optimization with transfer-based priors provides a scalable and robust framework for generating high-quality adversarial examples in challenging black-box scenarios.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Institute of Cyberspace Security, Zhejiang University of Technology, Hangzhou 310023, China.
- Binjiang Institute of Artificial Intelligence, Zhejiang University of Technology, Hangzhou 310056, China.
ORCID-Profile: Tianyi Ding, Xinjie Xu, Qi Xuan, Hanzhe Yu, Chen Ma
Zitieren
APA
Ding, T., Xu, X., Xuan, Q., Yu, H., & Ma, C. (2026). Query-Efficient Hard-Label Attack: A Prior-Guided Adam Ray Search Optimization. Sensors, 26(10), 3272. https://doi.org/10.3390/s26103272
Vancouver
Ding T, Xu X, Xuan Q, Yu H, Ma C. Query-Efficient Hard-Label Attack: A Prior-Guided Adam Ray Search Optimization. Sensors. 2026;26(10):3272. doi:10.3390/s26103272
BibTeX
@article{Ding2026,
author = {Ding, Tianyi and Xu, Xinjie and Xuan, Qi and Yu, Hanzhe and Ma, Chen},
title = {Query-Efficient Hard-Label Attack: A Prior-Guided Adam Ray Search Optimization},
journal = {Sensors},
year = {2026},
volume = {26},
number = {10},
pages = {3272},
doi = {10.3390/s26103272}
}
EUVIMED