Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network

Some recent articles have revealed that synthetic aperture radar automatic target recognition (SAR-ATR) models based on deep learning are vulnerable to the attacks of adversarial examples and cause security problems. The adversarial attack can make a deep convolutional neural network (CNN)-based SAR...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Chuan Du, Lei Zhang
Formato:	article
Lenguaje:	EN
Publicado:	MDPI AG 2021
Materias:	adversarial attack adversarial example generation UNet generative adversarial network (GAN) synthetic aperture radar (SAR) automatic target recognition (ATR) Science Q
Acceso en línea:	https://doaj.org/article/220f1be6a20243fb87864a46f691a522
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:220f1be6a20243fb87864a46f691a522
record_format	dspace
spelling	oai:doaj.org-article:220f1be6a20243fb87864a46f691a5222021-11-11T18:54:36ZAdversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network10.3390/rs132143582072-4292https://doaj.org/article/220f1be6a20243fb87864a46f691a5222021-10-01T00:00:00Zhttps://www.mdpi.com/2072-4292/13/21/4358https://doaj.org/toc/2072-4292Some recent articles have revealed that synthetic aperture radar automatic target recognition (SAR-ATR) models based on deep learning are vulnerable to the attacks of adversarial examples and cause security problems. The adversarial attack can make a deep convolutional neural network (CNN)-based SAR-ATR system output the intended wrong label predictions by adding small adversarial perturbations to the SAR images. The existing optimization-based adversarial attack methods generate adversarial examples by minimizing the mean-squared reconstruction error, causing smooth target edge and blurry weak scattering centers in SAR images. In this paper, we build a UNet-generative adversarial network (GAN) to refine the generation of the SAR-ATR models’ adversarial examples. The UNet learns the separable features of the targets and generates the adversarial examples of SAR images. The GAN makes the generated adversarial examples approximate to real SAR images (with sharp target edge and explicit weak scattering centers) and improves the generation efficiency. We carry out abundant experiments using the proposed adversarial attack algorithm to fool the SAR-ATR models based on several advanced CNNs, which are trained on the measured SAR images of the ground vehicle targets. The quantitative and qualitative results demonstrate the high-quality adversarial example generation and excellent attack effectiveness and efficiency improvement.Chuan DuLei ZhangMDPI AGarticleadversarial attackadversarial example generationUNetgenerative adversarial network (GAN)synthetic aperture radar (SAR)automatic target recognition (ATR)ScienceQENRemote Sensing, Vol 13, Iss 4358, p 4358 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	adversarial attack adversarial example generation UNet generative adversarial network (GAN) synthetic aperture radar (SAR) automatic target recognition (ATR) Science Q
spellingShingle	adversarial attack adversarial example generation UNet generative adversarial network (GAN) synthetic aperture radar (SAR) automatic target recognition (ATR) Science Q Chuan Du Lei Zhang Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network
description	Some recent articles have revealed that synthetic aperture radar automatic target recognition (SAR-ATR) models based on deep learning are vulnerable to the attacks of adversarial examples and cause security problems. The adversarial attack can make a deep convolutional neural network (CNN)-based SAR-ATR system output the intended wrong label predictions by adding small adversarial perturbations to the SAR images. The existing optimization-based adversarial attack methods generate adversarial examples by minimizing the mean-squared reconstruction error, causing smooth target edge and blurry weak scattering centers in SAR images. In this paper, we build a UNet-generative adversarial network (GAN) to refine the generation of the SAR-ATR models’ adversarial examples. The UNet learns the separable features of the targets and generates the adversarial examples of SAR images. The GAN makes the generated adversarial examples approximate to real SAR images (with sharp target edge and explicit weak scattering centers) and improves the generation efficiency. We carry out abundant experiments using the proposed adversarial attack algorithm to fool the SAR-ATR models based on several advanced CNNs, which are trained on the measured SAR images of the ground vehicle targets. The quantitative and qualitative results demonstrate the high-quality adversarial example generation and excellent attack effectiveness and efficiency improvement.
format	article
author	Chuan Du Lei Zhang
author_facet	Chuan Du Lei Zhang
author_sort	Chuan Du
title	Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network
title_short	Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network
title_full	Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network
title_fullStr	Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network
title_full_unstemmed	Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network
title_sort	adversarial attack for sar target recognition based on unet-generative adversarial network
publisher	MDPI AG
publishDate	2021
url	https://doaj.org/article/220f1be6a20243fb87864a46f691a522
work_keys_str_mv	AT chuandu adversarialattackforsartargetrecognitionbasedonunetgenerativeadversarialnetwork AT leizhang adversarialattackforsartargetrecognitionbasedonunetgenerativeadversarialnetwork
_version_	1718431631522398208

Adversarial Attack for SAR Target Recognition Based on UNet-Generative Adversarial Network

Ejemplares similares