SAR-ATR系统复数对抗样本生成方法
Plural Adversarial Sample Generation Method for SAR-ATR System
张梦君 1熊邦书1
作者信息
- 1. 南昌航空大学图像处理与模式识别江西省重点实验室,江西南昌 330063
- 折叠
摘要
针对现有对抗攻击方法只能用于攻击实数卷积神经网络这一限制,提出了一种基于生成对抗网络的复数对抗样本生成方法.首先,设计了一种产生有效对抗样本的复数模型,并引入了复数计算模块;其次,利用残差神经网络作为基本骨架,将预训练的复数网络作为判别器实现对抗训练,以增强对抗样本的攻击能力;最后,通过替代模型实现可迁移的对抗攻击,以此实现了更高的攻击成功率.实验结果表明,所提方法在有目标攻击和无目标攻击任务下的成功率分别达到了 76.338％和87.841％,迁移的成功率更高且对抗样本与原始干净样本更为接近.所提方法将对抗攻击扩展到复数神经网络后,避免了合成孔径雷达目标信息和精度的丢失,为实际合成孔径雷达自动目标识别系统的安全性和鲁棒性提供了参考方案.
Abstract
Existing adversarial attack methods are limited to real-valued convolutional neural networks.To address this limitation,this paper proposes a complex adversarial sample generation method based on generative adversarial networks.Firstly,a complex model for generating effective adversarial samples is designed by introducing complex com-putation modules.Secondly,a pre-trained complex network is used as the discriminator in adversarial training,with a residual neural network serving as the basic framework,to enhance the attack capability of adversarial samples.Finally,transferable adversarial attacks are achieved through substitute models,resulting in higher attack success rates.Experimental results demonstrate that the success rates of the proposed method in tar-geted and untargeted attack tasks reach 76.338％and 87.841％,respectively,with higher transferability and closer resemblance between adversarial and original clean samples.By extending adversarial attacks to complex neural networks,this method preserves synthetic aperture radar(SAR)target information and accuracy,providing a reference solution for the security and robustness of practical synthetic aperture radar automatic target recogni-tion(SAR-ATR)systems.
关键词
生成对抗网络/对抗样本/合成孔径雷达自动目标识别系统/复数卷积神经网络/有目标攻击/无目标攻击Key words
generative adversarial networks/adversarial examples/synthetic aperture radar automatic target recognition(SAR-ATR)/plural convolutional neural networks/tar-geted attack/untargeted attack引用本文复制引用
基金项目
国家自然科学基金(61866027)
江西省重点研发计划(20212BBE53017)
出版年
2024
NETL
NSTL
万方数据