Deep learning model security vulnerability testing and repairing method, device and system based on genetic algorithm

A deep learning and genetic algorithm technology, applied in the field of deep learning security, can solve the problem of less model testing work

Pending Publication Date: 2021-06-08
尚蝉(浙江)科技有限公司
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, there is less testing work for models subject to backdoor attacks,...

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Deep learning model security vulnerability testing and repairing method, device and system based on genetic algorithm
  • Deep learning model security vulnerability testing and repairing method, device and system based on genetic algorithm
  • Deep learning model security vulnerability testing and repairing method, device and system based on genetic algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0046] In order to realize security vulnerability detection of deep learning models such as automatic driving models or face recognition models, such as figure 1 As shown, the deep learning model security vulnerability testing method provided by the embodiment includes the following steps:

[0047] Step 1. Obtain the image dataset and the deep learning model to be tested.

[0048] In an embodiment, the image data set is an MNIST data set, an Imagenet data set or a Driving data set. The deep learning model is a LeNet deep learning model, a VGG19 deep learning model or a ResNet50 deep learning model.

[0049] Step 2, use the test deep learning model to test the images to filter images that can be correctly identified to form a clean image data set.

[0050] Specifically, the image in step S1 is input into the deep learning model to be tested, and the deep learning model for testing will output the predicted label of the input image. If the predicted label is consistent with th...

Embodiment 2

[0066] In order to repair the security vulnerabilities of deep learning models such as automatic driving models or face recognition models, such as figure 2 As shown, the genetic algorithm-based deep learning model security vulnerability repair method provided by the embodiment includes the following steps:

[0067] Step 1, using the above-mentioned genetic algorithm-based deep learning model security vulnerability testing method to test that there are security vulnerabilities in the deep learning model to be tested, and obtain the disturbed image as the test image;

[0068] Step 2, use the test image to optimize the training of the deep learning model to be tested, so as to repair the security holes of the deep learning model to be tested.

[0069] In the genetic algorithm-based deep learning model security vulnerability repair method provided in the embodiment, the obtained test image is used to perform intensive training on the original deep learning model to repair the de...

Embodiment 3

[0071] In order to realize security vulnerability detection of deep learning models such as automatic driving models or face recognition models, such as image 3 As shown, the deep learning model security vulnerability testing device 300 provided by the embodiment includes:

[0072] The building block 301 is used to obtain the image data set and the deep learning model to be tested, and use the test deep learning model to test the image to filter images that can be correctly identified to form a clean image data set;

[0073] Screening module 302, for randomly selecting some images from the clean image data set as test seed images, and adding initial perturbation to the test seed images;

[0074] The detection module 303 is used to input the disturbed image to the deep learning model to be tested to obtain the predicted label, select the image as the parent according to the fitness function constructed by minimizing the added disturbance and the difference between the predicte...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a deep learning model security vulnerability testing and repairing method, device and system based on a genetic algorithm. An evaluation function is constructed according to the class prediction value difference of a label of an input image in a deep learning model and the added noise minimization, the evaluation function is optimized to generate a malignant image which can cover the boundary of the deep learning model and has a large identification gap, when the malignant image can be obtained, it shows that the original deep learning model is unsafe, otherwise, the original deep learning model is safe, so that the security test of the deep learning model is realized, and the security test method is simple and accurate; and the obtained test image is used to carry out intensified training on the original deep learning model so as to repair the deep learning model and improve the recognition result accuracy of the deep learning model.

Description

technical field [0001] The invention belongs to the technical field of deep learning security, and in particular relates to a genetic algorithm-based deep learning model security loophole testing and repair method, device and system. Background technique [0002] Artificial intelligence technology has made breakthroughs in fields such as computer vision and natural language processing, ushering in a new round of explosive development of artificial intelligence. As the key technology in these breakthroughs, deep learning has gradually become a research hotspot and mainstream development direction in the field of artificial intelligence. Deep learning is a computational model composed of multiple processing layers, a machine learning technique that learns data representations with multiple levels of abstraction. Deep learning represents the main development direction of machine learning and artificial intelligence research, and has brought revolutionary progress to the fields...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
IPC IPC(8): G06F21/55G06F21/57G06K9/62G06N3/04G06N3/08
CPCG06F21/554G06F21/577G06N3/086G06N3/045G06F18/241G06F18/214
Inventor 纪守领林昶廷董建锋王睿
Owner 尚蝉(浙江)科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap