Supercharge Your Innovation With Domain-Expert AI Agents!

Agile image encryption method based on deep learning

A technology of deep learning and encryption method, which is applied in the field of image information security, can solve problems such as system inconvenience, and achieve the effect of enhancing security and being convenient and quick to use

Active Publication Date: 2022-01-28
ZHEJIANG GONGSHANG UNIVERSITY
View PDF12 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, using network parameters as keys has a relatively large disadvantage: if users want to update the keys, they need to retrain a new network, and they need to keep their own encryption and decryption networks secret and maintain. The system is very inconvenient

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
  • Agile image encryption method based on deep learning
  • Agile image encryption method based on deep learning
  • Agile image encryption method based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0074] (1) Prepare training data.

[0075] Prepare a sufficient number of training samples (the training set data can be collected and created by yourself or downloaded from the Internet to some public image data sets, such as ImageNet, etc.), and all images are converted into grayscale images, and all images are normalized to a size of 256×256 , the normalization method is as follows: For images whose short side is greater than 256, directly perform random cropping; for images whose short side is less than 256, first enlarge the image proportionally until the short side is 256, and then perform random cropping.

[0076] (2) Joint training encryption network and decryption network

[0077] Input the prepared training data into the encryption network and the decryption network in batches for training. For each batch of data, first randomly generate two different keys k≠k', and use the key as a seed to generate a pseudo key containing 32 integers. Random sequence, each integer ...

Embodiment 2

[0083] (1) Prepare training data.

[0084] Prepare a sufficient number of training samples (the training set data can be collected and created by yourself or downloaded from the Internet to some public image data sets, such as ImageNet, etc.), and all images are converted into grayscale images, and all images are normalized to a size of 256×256 , the normalization method is as follows: For images whose short side is greater than 256, directly perform random cropping; for images whose short side is less than 256, first enlarge the image proportionally until the short side is 256, and then perform random cropping.

[0085] (2) Joint training encryption network and decryption network

[0086] Input the prepared training data into the encryption network and the decryption network in batches for training. For each batch of data, first randomly generate two different keys k≠k', and use the key as a seed to generate a pseudo key containing 32 integers. Random sequence, each integer ...

Embodiment 3

[0092] (1) Prepare training data.

[0093] Prepare a sufficient number of training samples (the training set data can be collected and created by yourself or downloaded from the Internet to some public image data sets, such as ImageNet, etc.), and all images are converted into grayscale images, and all images are normalized to a size of 512×512 , the normalization method is as follows: for images whose short side is larger than 512, directly perform random cropping; for images whose short side is smaller than 512, first enlarge the image proportionally until the short side is 512, and then perform random cropping.

[0094] (2) Joint training encryption network and decryption network

[0095] Input the prepared training data into the encryption network and the decryption network in batches for training. For each batch of data, first randomly generate two different keys k≠k', and use the key as a seed to generate a pseudo key containing 32 integers. Random sequence, each intege...

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 an agile image encryption method based on deep learning. An encryption network and a decryption network utilize an improved ResNet as a backbone network, all pooling layers in the ResNet are removed, hole convolution is introduced into residual connection of a residual module, and a multi-scale fusion structure is adopted. Meanwhile, in order to improve the security of the image encryption method, the encrypted image based on the deep learning convolutional neural network is further confused, so that the key space of the encryption method is greatly expanded, the avalanche effect of the image encryption method is enhanced, and the difficulty of group search attack and differential attack is improved. The image encryption method is easy to integrate with an artificial intelligence system based on deep learning, so that the privacy protection capability of the artificial intelligence system is improved. The method has better time performance, the encryption and decryption network does not need to be retrained when the key is modified, the model structure and parameters can be disclosed, the method conforms to the Kerckhoffs's principle, and the method is convenient and fast to use.

Description

technical field [0001] The invention relates to the technical field of image information security, in particular to an agile image encryption method based on deep learning. Background technique [0002] With the development of network technology and media recording equipment, more and more multimedia information is transmitted or shared through the network and storage devices. Certain images contain personal sensitive information, and people face privacy protection issues when they do not want these images to be accessed by unauthorized persons. Biological information images such as face images, iris images, and fingerprint images are widely used as identity identification information in the security field and are personal sensitive information; in addition, medical images also contain private information that needs to be protected, and military images are confidential information that needs to be protected. Protect. Modern artificial intelligence systems require a large n...

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
Patent Type & Authority Applications(China)
IPC IPC(8): H04N1/44H04L9/00H04L9/08G06N3/04G06N3/08
CPCH04N1/44H04L9/001H04L9/0869G06N3/084H04L2209/08G06N3/048G06N3/045
Inventor 竺乐庆马佳琪
Owner ZHEJIANG GONGSHANG UNIVERSITY
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com