Adversarial sample attack detection method and device based on deep learning, and electronic equipment

Abstract

The invention provides an adversarial sample attack detection method and device based on deep learning and electronic equipment, and the method comprises the steps: generating an adversarial sample corresponding to a clean sample through employing an optimized mode based on a momentum iteration gradient; and determining an anti-attack performance evaluation result of the deep neural network model according to the adversarial sample corresponding to the clean sample. According to the method, the anti-attack performance of the deep neural network model can be evaluated more effectively.

Description

technical field [0001] The present application relates to the field of artificial intelligence security technology, and in particular, to a method, device and electronic device for detecting an adversarial sample attack based on deep learning. Background technique [0002] At present, the development of deep learning provides a reliable tool for the application of artificial intelligence. Deep Neural Networks (DNNs) have been widely used in computer vision (such as face recognition, target detection, autonomous driving) and other fields. . On natural images (such as the CIFAR-10 and ImageNet datasets), state-of-the-art convolutional neural networks have outperformed the human eye in image classification tasks. Due to the excellent performance of DNNs and the high cost of traditional tools, replacing them with deep learning algorithms is a suitable choice, so DNNs have also become popular tools for image processing tasks. [0003] Despite the superior performance of DNNs, t...

AI Technical Summary

Problems solved by technology

[0003] Despite the superior performance of DNNs, studies have found that DNNs are vulnerable to well-designed adversarial samples, which can be generated by adding small perturbations that cannot be recognized by the naked eye in the original image, that is, slightly perturbed input samples can make DNNs make wrong prediction

The existence of adversarial examples makes the application of DNNs in the image domain worrying

Images