High-transferability confrontation sample generation method, system and terminal

Abstract

The invention discloses a high-transitivity adversarial sample generation method and system, and a terminal, and belongs to the technical field of deep learning, and the method comprises the steps: adding the low-frequency information of a first target image to a randomly selected original image, and obtaining a first sample; randomly selecting a primary sampling model from the model pool; inputting the first sample into a primary sampling model; iteratively calculating an aggregation gradient of the sampling model according to low-frequency information of a first target image added to the original image, and normalizing the aggregation gradient; updating an adversarial sample generated by the sampling model based on a gradient symbol, constraining an adversarial sample value, and outputting a first-level temporary adversarial sample as input of a lower-level sampling model; and repeating the steps to obtain a final generated sample. The method is provided with a multi-stage sampling model, and the transferability of the finally generated adversarial sample is improved. And meanwhile, target low-frequency information different from the category of the sample is superposed on the sample input into the sampling model, so that the transferability of the adversarial sample can be further improved.

Description

technical field [0001] The present invention relates to the technical field of deep learning, in particular to a method, system and terminal for generating highly transferable adversarial samples. Background technique [0002] In recent years, with the rapid development of deep learning, its intelligent applications in image recognition, face recognition, unmanned driving, target detection, medical treatment and other fields have become a reality. Compared with traditional methods, deep learning has shown extremely high performance in various fields. However, some studies have shown that deep learning is vulnerable to adversarial attacks due to the existence of adversarial samples. In today's highly informatized and intelligent world, the security guarantee of deep learning algorithms is an unprecedented difficulty and pain point. [0003] The adversarial samples generated by the attacker through carefully designed perturbations are indistinguishable from the real samples t...

AI Technical Summary

Problems solved by technology

[0008] The purpose of the present invention is to overcome the problem of low transferability of adversarial samples generated in the prior art in model attacks, and provide a method, system and terminal for generating adversarial samples with high transferability

Images