Abstract:Recent advances in deep neural networks (DNNs) highlight the success of synthetic aperture radar automatic target recognition (SAR ATR) with superior effectiveness and efficiency. However, the DNNs are known to be vulnerable to adversarial examples, whose performance will be dramatically reduced when the imperceptible perturbation exists. In optical image processing, invisible perturbations are typically embedded in the way of a full-scale distribution in a purely digital setting. Whereas, it is not feasible to achieve this in SAR ATR tasks due to the inaccessibility of the SAR system and unique imaging mechanism. In practice, the subtle perturbations could be produced by physical approaches that change the scattering property of the target. Therefore, the adversarial perturbations for SAR ATR should be of good transferability to achieve an effective attack on major DNN classifiers as well as the accessible additive region in SAR images with respect to the realistic target locations. In this letter, we present a novel approach, namely speckle-variant attack (SVA). The proposed SVA is composed of two major modules: an iterative gradient-based perturbation generator and a target region extractor. The perturbation generator implements a speckle-variant transformation that continuously reconstructs the speckle-noise pattern during each of the iterations for strong transferability. The target region extractor ensures the feasibility of the additive adversarial perturbations in practical scenarios by restricting the region of the perturbation. Therefore, the proposed SVA is capable of producing adversarial examples that are more transferable and physically feasible. Extensive evaluations of the MSTAR dataset show that the SVA has achieved superior transferability and competitive time consumption compared with the SOTA transformation-based techniques, including the diverse inputs method and the scale-invariant method.

Fast C&W: A Fast Adversarial Attack Algorithm to Fool SAR Target Recognition With Deep Convolutional Neural Networks

Fast sparse adversarial attack for synthetic aperture radar target recognition

A Novel Group Squeeze Excitation Sparsely Connected Convolutional Networks for SAR Target Classification

SCMA: A Scattering Center Model Attack on CNN-SAR Target Recognition

DNN-Based Peak Sequence Classification CFAR Detection Algorithm for High-Resolution FMCW Radar

Adversarial Examples for CNN-Based SAR Image Classification: an Experience Study

A Universal Adversarial Attack on CNN-SAR Image Classification by Feature Dictionary Modeling.

Speckle-Variant Attack: Toward Transferable Adversarial Attack to SAR Target Recognition

Adversarial Examples of SAR Images for Deep Learning Based Automatic Target Recognition

Convolutional Neural Network With Attention Mechanism for SAR Automatic Target Recognition

Scattering Model Guided Adversarial Examples for SAR Target Recognition: Attack and Defense

Black-Box Universal Adversarial Attack for DNN-Based Models of SAR Automatic Target Recognition

TAN: A Transferable Adversarial Network for DNN-Based UAV SAR Automatic Target Recognition Models

Learning Invariant Representation Via Contrastive Feature Alignment for Clutter Robust SAR ATR

Learning Invariant Representation via Contrastive Feature Alignment for Clutter Robust SAR Target Recognition

SAR-AE-SFP: SAR Imagery Adversarial Example in Real Physics domain with Target Scattering Feature Parameters

A Universal Defense Strategy Against Adversarial Attacks Based on Attention-Guided

Boosting Adversarial Transferability with Shallow-Feature Attack on SAR Images

Learning to Generate SAR Images With Adversarial Autoencoder

A New Algorithm for SAR Image Target Recognition Based on an Improved Deep Convolutional Neural Network

FCCD-SAR: A Lightweight SAR ATR Algorithm Based on FasterNet