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

Abstract

Deep learning technology (a deeper and optimized network structure) and remote sensing imaging (i.e., the more multisource and the more multicategory remote sensing data) have developed rapidly. Although the deep convolutional neural network (CNN) has achieved state-of-the-art performance on remote sensing image (RSI) scene classification, the existence of adversarial attacks poses a potential security threat to the RSI scene classification task based on CNN. The corresponding adversarial samples can be generated by adding a small perturbation to the original images. Feeding the CNN-based classifier with the adversarial samples leads to the classifier misclassify with high confidence. To achieve a higher attack success rate against scene classification based on CNN, we introduce the projected gradient descent method to generate adversarial remote sensing images. Then, we select several mainstream CNN-based classifiers as the attacked models to demonstrate the effectiveness of our method. The experimental results show that our proposed method can dramatically reduce the classification accuracy under untargeted and targeted attacks. Furthermore, we also evaluate the quality of the generated adversarial images by visual and quantitative comparisons. The results show that our method can generate the imperceptible adversarial samples and has a stronger attack ability for the RSI scene classification.

Defensive Distillation-Based Adversarial Attack Mitigation Method for Channel Estimation Using Deep Learning Models in Next-Generation Wireless Networks

“…PGD is one of the most popular and powerful attacks, which is called gradient-based attacks [40], [41]. It is used to compute the gradient of the loss function with respect to the input, x, and then the attacker creates the adversarial example by adding the sign of the gradient to the input data.…”

The Adversarial Security Mitigations of mmWave Beamforming Prediction Models using Defensive Distillation and Adversarial Retraining

“…However, it has a different method to generate adversarial examples. It initials the search for the adversarial example at random points in a suitable region, then runs several iterations to find an adversarial example with the greatest loss, but the size of the perturbation is smaller than a specified amount referred to as epsilon, [20]. PGD can generate stronger attacks than FGSM and BIM.…”

Defensive Distillation based Adversarial Attacks Mitigation Method for Channel Estimation using Deep Learning Models in Next-Generation Wireless Networks

“…The BIM attack has been used in [32] to attack channel estiomation prediction models. 3) PGD: PGD is one of the most popular attacks, which is called gradient-based attacks [33]. It is called the most powerful attack.…”