Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Guo, Zihong, Wan, Chen, Zheng, Yayin, Kuang, Hailing, Lu, Xiaohai
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2507.01791
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915369594126336
author Guo, Zihong
Wan, Chen
Zheng, Yayin
Kuang, Hailing
Lu, Xiaohai
author_facet Guo, Zihong
Wan, Chen
Zheng, Yayin
Kuang, Hailing
Lu, Xiaohai
contents The transferability of adversarial examples poses a significant security challenge for deep neural networks, which can be attacked without knowing anything about them. In this paper, we propose a new Segmented Gaussian Pyramid (SGP) attack method to enhance the transferability, particularly against defense models. Unlike existing methods that generally focus on single-scale images, our approach employs Gaussian filtering and three types of downsampling to construct a series of multi-scale examples. Then, the gradients of the loss function with respect to each scale are computed, and their average is used to determine the adversarial perturbations. The proposed SGP can be considered an input transformation with high extensibility that is easily integrated into most existing adversarial attacks. Extensive experiments demonstrate that in contrast to the state-of-the-art methods, SGP significantly enhances attack success rates against black-box defense models, with average attack success rates increasing by 2.3% to 32.6%, based only on transferability.
format Preprint
id arxiv_https___arxiv_org_abs_2507_01791
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Boosting Adversarial Transferability Against Defenses via Multi-Scale Transformation
Guo, Zihong
Wan, Chen
Zheng, Yayin
Kuang, Hailing
Lu, Xiaohai
Computer Vision and Pattern Recognition
The transferability of adversarial examples poses a significant security challenge for deep neural networks, which can be attacked without knowing anything about them. In this paper, we propose a new Segmented Gaussian Pyramid (SGP) attack method to enhance the transferability, particularly against defense models. Unlike existing methods that generally focus on single-scale images, our approach employs Gaussian filtering and three types of downsampling to construct a series of multi-scale examples. Then, the gradients of the loss function with respect to each scale are computed, and their average is used to determine the adversarial perturbations. The proposed SGP can be considered an input transformation with high extensibility that is easily integrated into most existing adversarial attacks. Extensive experiments demonstrate that in contrast to the state-of-the-art methods, SGP significantly enhances attack success rates against black-box defense models, with average attack success rates increasing by 2.3% to 32.6%, based only on transferability.
title Boosting Adversarial Transferability Against Defenses via Multi-Scale Transformation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2507.01791