基于掩模提取的SAR图像对抗样本生成方法

doi:10.11959/j.issn.1000-0801.2024081

Abstract

Abstract:

There are many ways to generate adversarial samples for synthetic aperture radar (SAR) images at present, but some problems such as large amount of perturbation of adversarial samples, unstable training, and unguaranteed quality of adversarial samples still exist.To solve the above problems, a SAR image adversarial sample generation model was proposed.The model was based on the AdvGAN model architecture.Firstly, according to the characteristics of the SAR images, an adaptive threshold segmentation method based on the enhanced Lee filter OTSU was designed.The mask extraction module composed of equal modules, this method produced a smaller amount of disturbance, and the structural similarity (SSIM) with the original sample reached that more than 0.997.Secondly, the improved relativistic average GAN (RaGAN) loss was introduced into AdvGAN, and the relative mean discriminator was used to make the discriminator rely on both real data and generated data during training, which improved the stability of training and the attack effect.Experiments were compared with related methods on the MSTAR dataset.Experiments show that the attack success rate of SAR image adversarial samples generated by this method is increased by 10%～15% than that of traditional methods when attacking defense models.

Key words: adversarial sample, generative adversarial network, synthetic aperture radar, semi-white box attack, mask extraction

CLC Number:

TN957

Jianwu ZHANG, Hao NAI, Jie LI, Jianhua QIAN, Yinfeng FANG. Adversarial example generation method for SAR images based on mask extraction[J]. Telecommunications Science, 2024, 40(3): 64-74.

Figures/Tables 10

目标模型	攻击模型	攻击成功率	SSIM
AlexNet	AdvGAN	92.80%	0.980 6
	Ra_AdvGAN	$96 . 30 %$	$0 . 9869$
VggNet16	AdvGAN	94.59%	0.939 3
	Ra_AdvGAN	$98 . 15 %$	$0 . 9488$
ResNet18	AdvGAN	92.15%	0.974 2
	Ra_AdvGAN	$97 . 82 %$	$0 . 9766$
ResNet50	AdvGAN	92.17%	0.962 4
	Ra_AdvGAN	$96 . 89 %$	$0 . 9806$
DenseNet121	AdvGAN	95.27%	0.744 0
	Ra_AdvGAN	$98 . 60 %$	$0 . 7923$

攻击模型	目标模型	攻击成功率	SSIM
AdvGAN	VggNet16	94.59%	0.939 3
	ResNet50	92.17%	0.962 4
	防御模型	67.05%	—
基于CycleGAN	VggNet16	94.51%	0.965 4
	ResNet50	93.01%	0.963 7
	防御模型	68.97%	—
基于WGAN	VggNet16	95.39%	0.968 8
	ResNet50	95.44%	0.957 9
	防御模型	74.44%	—
Ra_AdvGAN	VggNet16	$98 . 15 %$	0.976 6
	ResNet50	$96 . 89 %$	0.980 6
	防御模型	74.17%	—
基于Ra_AdvGAN和	VggNet16	86.21%	$0 . 9980$
掩模提取	ResNet50	87.19%	$0 . 9976$
	防御模型	$83 . 11 %$	—

References 18

[1]	XIA W J , LIU Z , LI Y . SAR-PeGA:a generation method of adversarial examples for SAR image target recognition network[J]. IEEE Transactions on Aerospace and Electronic Systems, 2023,59(2): 1910-1920.
[2]	何涛, 李大亮, 曹兰英 . 基于深度学习的机载SAR典型目标识别算法[J]. 现代雷达, 2022,44(12): 87-92.
	HE T , LI D L , CAO L Y . An algorithm of typical target recognition for airborne SAR based on deep learning[J]. Modern Radar, 2022,44(12): 87-92.
[3]	SHI J . SAR target recognition method of MSTAR data set based on multi-feature fusion[C]// Proceedings of the 2022 International Conference on Big Data,Information and Computer Network (BDICN). Piscataway:IEEE Press, 2022: 626-632.
[4]	FANG C , SONG Y M , GUAN F H ,et al. A robust complexvalued deep neural network for target recognition of UAV SAR imagery[J]. IEEE Journal on Miniaturization for Air and Space Systems, 2023,4(2): 175-185.
[5]	王志波, 王雪, 马菁菁 ,等. 面向计算机视觉系统的对抗样本攻击综述[J]. 计算机学报, 2023,46(2): 436-468.
	WANG Z B , WANG X , MA J J ,et al. Survey on adversarial example attack for computer vision systems[J]. Chinese Journal of Computers, 2023,46(2): 436-468.
[6]	赵宏, 常有康, 王伟杰 . 深度神经网络的对抗攻击及防御方法综述[J]. 计算机科学, 2022,49(S2): 662-672.
	ZHAO H , CHANG Y K , WANG W J . Summary of anti-attack and defense methods of deep neural networks[J]. Computer Science, 2022,49(S2): 662-672.
[7]	WANG X M , LI J , KUANG X H ,et al. The security of machine learning in an adversarial setting:a survey[J]. Journal of Parallel and Distributed Computing, 2019(130): 12-23.
[8]	HUANG T , ZHANG Q X , LIU J B ,et al. Adversarial attacks on deep-learning-based SAR image target recognition[J]. Journal of Network and Computer Applications, 2020(162): 102632.
[9]	MOOSAVI-DEZFOOLI S M , FAWZI A , FROSSARD P . DeepFool:a simple and accurate method to fool deep neural networks[C]// Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2016: 2574-2582.
[10]	CARLINI N , WAGNER D . Towards evaluating the robustness of neural networks[C]// Proceedings of the 2017 IEEE Symposium on Security and Privacy (SP). Piscataway:IEEE Press, 2017: 39-57.
[11]	XIAO C W , LI B , ZHU J Y ,et al. Generating adversarial examples with adversarial networks[C]// Proceedings of the Proceedings of the 27th International Joint Conference on Artificial Intelligence. New York:ACM Press, 2018: 3905-3911.
[12]	LIU X Q , HSIEH C J . Rob-GAN:generator,discriminator,and adversarial attacker[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE Press, 2019: 11226-11235.
[13]	张高志, 刘新平, 邵明文 . 用于白盒目标攻击的GAN对抗样本生成[J]. 模式识别与人工智能, 2020,33(9): 830-838.
	ZHANG G Z , LIU X P , SHAO M W . Generating adversarial example with GAN for white-box target attacks[J]. Pattern Recognition and Artificial Intelligence, 2020,33(9): 830-838.
[14]	田宇 . 基于GAN的图像对抗样本生成方法研究[D]. 北京:北京邮电大学, 2020.
	TIAN Y . Research of method on generating image adversarial samples based on GAN[D]. Beijing:Beijing University of Posts and Telecommunications, 2020.
[15]	付睢宁, 路泽忠, 王舜瑶 . 一种改进的Lee滤波SAR图像去噪算法[J]. 计算机与数字工程, 2019,47(8): 2018-2021.
	FU S N , LU Z Z , WANG S Y . An improved lee filter SAR image denoising algorithm[J]. Computer ＆ Digital Engineering, 2019,47(8): 2018-2021.
[16]	谢鹏鹤 . 图像阈值分割算法研究[D]. 湘潭:湘潭大学, 2012.
	XIE P H . The study on the image thresholding segmentation algorithm[D]. Xiangtan:Xiangtan University, 2012.
[17]	JOLICOEUR-MARTINEAU A . The relativistic discriminator:a key element missing from standard GAN[EB]. 2018:arXiv:1807.00734.
[18]	CARLINI N , WAGNER D . Towards evaluating the robustness of neural networks[C]// Proceedings of the 2017 IEEE Symposium on Security and Privacy (SP). Piscataway:IEEE Press, 2017: 39-57.

Metrics

Recommended 0

No Suggested Reading articles found!

特性	光学图像	SAR图像
波长	纳米级	厘米级
应用	广泛应用于各领域	军用
数据形式	RGB三通道	单通道复数
分辨率	高	低
样本数量	多	少
成像原理	能量聚焦积累	相位相干积累

Adversarial example generation method for SAR images based on mask extraction

RichHTML

PDF下载

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 18

Related Articles 7

Metrics

Recommended 0

[1]	Yuanzhi LU, Xianglin WEI, Long YU, Changhua YAO. 5G OFDM channel estimation method based on complexvalued generative adversarial network [J]. Telecommunications Science, 2024, 40(3): 39-52.
[2]	Hui MA, Ruiqin WANG, Shuai YANG. A progressive growing of conditional generative adversarial networks model [J]. Telecommunications Science, 2023, 39(6): 105-113.
[3]	Kehong A, Xiaodong HU. GAN data reconstruction based prediction method of telecom subscriber loss [J]. Telecommunications Science, 2023, 39(3): 135-142.
[4]	Jinyang SUN, Baisong LIU, Hao REN, Jiangbo QIAN. TAGAN: an academic paper adversarial recommendation algorithm incorporating fine-grained semantic features [J]. Telecommunications Science, 2021, 37(8): 57-65.
[5]	Yuxiu HUA, Rongpeng LI, Zhifeng ZHAO, Jianjun WU, Honggang ZHANG. GAN-based channel estimation for massive MIMO system [J]. Telecommunications Science, 2021, 37(6): 14-22.
[6]	Shengsheng ZHENG, Haibing YIN, Xiaofeng HUANG, Tianjie ZHANG. GAN-based unsupervised domain adaptive person re-identification [J]. Telecommunications Science, 2021, 37(2): 99-106.
[7]	Yaguan QIAN,Xiaohui GUAN,Shuhui WU,NBensheng YU,Dongxiao REN. A novel perturbation attack on SVM by greedy algorithm [J]. Telecommunications Science, 2019, 35(1): 81-89.