Abstract:Despite deep neural networks (DNNs) have been widely applied in remote sensing (RS) scene classification and achieved satisfying performance, the vulnerability of DNNs toward adversarial examples significantly degrades their performance. Moreover, the relatively limited labeled samples of RS scene classification make DNNs more likely to overfit, leading to weak generalizability and noise sensitivity. This may result in DNNs being more vulnerable to adversarial examples. Consequently, the defense of adversarial examples is of crucial importance to improve both the generalizability and robustness of DNNs in the RS scene classification task. However, few studies have been conducted on defense for RS scene classification, especially ignoring the intrinsic characteristics of RS images. In this article, an effective defense framework for RS scene classification, named reconstruction-assisted and distance-optimized adversarial training (RDAT), is proposed to defend adversarial examples. To solve the problems caused by high interclass similarity, a distance-optimized (DO) strategy is designed for adversarial training (AT) to strengthen the learning of underfitting content, increase the interclass distance, and improve the robustness of the networks. Furthermore, to generate high-quality samples for AT, a reconstruction-assisted (RA) block is proposed to eliminate adversarial perturbations in adversarial examples. Specifically, in this block, by Swin Transformer (SwinT) block and multiscale convolution (MSC) block, SwinT-MSC-UNet (SMUNet) is constructed to fully extract global and multiscale local features to adapt to the characteristics of RS images with a large variance of ground object scales. Extensive experiments on the benchmark datasets, that is, UC Merced (UCM) and aerial image dataset (AID), have demonstrated that the proposed RDAT can effectively resist multiple adversarial attacks and yield superior results than other defense methods for RS scene classification.

Project Gradient Descent Adversarial Attack Against Multisource Remote Sensing Image Scene Classification

Reconstruction-Assisted and Distance-Optimized Adversarial Training: A Defense Framework for Remote Sensing Scene Classification

Attack Selectivity Of Adversarial Examples In Remote Sensing Image Scene Classification

Generating Adversarial Examples Against Remote Sensing Scene Classification Via Feature Approximation

Assessing the Threat of Adversarial Examples on Deep Neural Networks for Remote Sensing Scene Classification: Attacks and Defenses

Feature Decoupling Based Adversarial Examples Detection Method for Remote Sensing Scene Classification

An Adversarial Attack Via Feature Contributive Regions

CamoNet: A Target Camouflage Network for Remote Sensing Images Based on Adversarial Attack

An Empirical Study of Adversarial Examples on Remote Sensing Image Scene Classification

Multifeature Collaborative Adversarial Attack in Multimodal Remote Sensing Image Classification.

Adversarial Example in Remote Sensing Image Recognition

FCGSM: Fast Conjugate Gradient Sign Method for Adversarial Attack on Image Classification

Adversarial Example Generation Method for Object Detection in Remote Sensing Images

Adversarial Defense Method Based on Latent Representation Guidance for Remote Sensing Image Scene Classification.

Perturbation-Seeking Generative Adversarial Networks: A Defense Framework for Remote Sensing Image Scene Classification

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

Lie to Me: A Soft Threshold Defense Method for Adversarial Examples of Remote Sensing Images

Multi-patch Adversarial Attack for Remote Sensing Image Classification

Robust Remote Sensing Scene Interpretation Based on Unsupervised Domain Adaptation

Target Attack on Biomedical Image Segmentation Model Based on Multi-Scale Gradients.

Dual-Branch Sparse Self-Learning with Instance Binding Augmentation for Adversarial Detection in Remote Sensing Images