2021 the 5th International Conference on Innovation in Artificial Intelligence | 2021
Adversarial Examples Generation And Attack On SAR Image Classification
Abstract
It has been demonstrated that deep learning models are vulnerable to adversarial examples, and most existing algorithms can generate adversarial examples to attack image classification or recognition models trained from target datasets with visible image such as ImageNet, PASCAL VOC and COCO. In order to expand the image style of adversarial examples and decide whether the attacking specialty exists in SAR images, the MI-FGSM and AdvGAN algorithms are introduced to generate adversarial examples, and the attacks for SAR image classification are executed in this paper. The experimental results show that the adversarial example is also definitely aggressive to SAR image, and the attack success rate is affected by the structure of deep learning networks training the target models and the category of the image.