Information processing method and device, medium and computing equipment

Abstract

The embodiment of the invention provides an information processing method and device, a storage medium and computing equipment. The information processing method comprises the steps of obtaining a predetermined image recognition model; for the image recognition model, based on an input-output relationship obtained by accessing an image recognition model by using candidate confrontation samples searched in an input space each time, solving an optimal solution of a preset optimization problem through iterative calculation of gradually reducing the distance between the candidate confrontation sample and the original sample, wherein the preset optimization problem comprises the steps of obtaining the candidate confrontation sample with the minimum distance from the original sample in an inputspace; and enabling the candidate confrontation samples searched in each iteration to successfully attack the image recognition model. According to the technology provided by the invention, the imagerecognition model given for resisting the successful attack of the sample can be generated under the condition that the model structure and parameters are not obtained, the attack effect on the face recognition model is especially good, and compared with the prior art, the access frequency can be greatly reduced.

Description

technical field [0001] Embodiments of the present invention relate to the field of information processing, and more specifically, embodiments of the present invention relate to an information processing method and device, a medium, and a computing device. Background technique [0002] As a class of machine learning methods, deep neural networks have gained widespread attention in recent years due to their remarkable results in many fields such as speech recognition, image classification, and object detection. However, deep neural network models that can achieve high accuracy on many tasks are vulnerable to attacks in adversarial environments. In the adversarial environment, network models such as deep neural networks will be input with some maliciously constructed adversarial samples based on normal samples, such as pictures or voice information. These adversarial examples are easily misclassified by deep learning models, but it is difficult for human observers to detect th...

AI Technical Summary

Problems solved by technology

Although face recognition systems have achieved great success and practical applications, the security of these systems has not received enough attention

Images