Intelligent adversarial attack sample generation method and system based on scaling transformation

Abstract

The present application relates to the technical field of image processing, and particularly relates to a kind of intelligent adversarial attack sample generation method and system based on scaling transformation, obtain original image data and its corresponding prior label;Add disturbance in original image data based on iterative optimization method, to generate adversarial sample and output, wherein, in each iteration, the adversarial sample image data generated in the last iteration is processed multiple times by scaling expansion, the image data after scaling expansion is input into network model and the gradient of target loss function is calculated using gradient optimization method, the disturbance in the current iteration is obtained according to the gradient of target loss function, and the disturbance is added to the adversarial sample generated in the last iteration to generate the adversarial sample of the current iteration.The present application can effectively expand the training set and reduce the overfitting in the process of generating adversarial samples, improve the migration of adversarial samples and the success rate of black box attack, improve the quality of adversarial sample generation, facilitate image classification, target detection, face recognition and other scene applications.

Images