Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Deep learning differential privacy protection method

A differential privacy and deep learning technology, applied in the field of information system security, can solve problems such as parameter shocks, and achieve the effects of improving usability, improving practical significance, and high data availability

Pending Publication Date: 2020-10-02
ANHUI UNIVERSITY OF TECHNOLOGY
View PDF2 Cites 7 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the DCGAN used in this method still has insufficient training stability. As the number of training increases, some parameters (such as filter) will oscillate due to collapse, and the generation model of DCGAN will be limited by batch normalization.
In addition, the group setting method of privacy parameters in this method mainly depends on the individual needs of users, and there is no qualitative analysis of privacy loss minimization for feedback-adjusted privacy parameter settings

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 differential privacy protection method
  • Deep learning differential privacy protection method
  • Deep learning differential privacy protection method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0051] to combine figure 1 , a deep learning differential privacy protection method of the present embodiment, the steps of which are:

[0052] Step 1. Construct a convolutional neural network with two convolutional layers and three fully connected layers, introduce differential privacy theory into network parameter optimization, and add Gaussian noise that satisfies the Gaussian mechanism. The specific process is as follows:

[0053] Initialize, establish a convolutional neural network with two convolutional layers and three fully connected layers, and initialize the model parameters of the convolutional neural network, such as image 3 shown. Introduce (ε,δ)-differential privacy as shown in formula (1):

[0054] Pr[M(D)∈S M ]≤e ε ×Pr[M(D’)∈S M ]+δ (1)

[0055] Among them, M is a given random algorithm; D and D' are neighbor data sets with a difference of at most one record; S M are all possible outputs of random algorithm M on data sets D and D'. The degree of privac...

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 differential privacy protection method, and belongs to the technical field of information system security. The invention provides a novel deep learning differential privacy protection model. A WGAN is adopted to generate an image result for the data subjected to model privacy protection processing; a result closest to a real image is selected from the generated images, the similarity between the generated result and the original image is compared, a difference value is calculated for threshold comparison, privacy parameters in the model gradient are fedback and adjusted under the similarity threshold limiting condition, and therefore a certain promotion effect is provided for application of differential privacy in the fields of deep learning and thelike.

Description

technical field [0001] The invention belongs to the technical field of information system security, and more specifically relates to a deep learning differential privacy protection method. Background technique [0002] The existing privacy protection methods for relatively common data sets, such as using k-anonymity to anonymize data, etc., are difficult to provide strict privacy guarantees after the actual processing effect. As a new type of privacy protection technology with great advantages, Differential privacy (DR) technology is a privacy protection method based on data distortion proposed for attackers with strong knowledge background, by adding noise to ensure Inserting or deleting any record in the data set does not affect the query output results, so as to achieve the purpose of protecting data privacy. This technology is based on a rigorous mathematical foundation and provides a quantitative evaluation method. It is one of the most effective and highly applicable ...

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): G06F21/62G06N3/04G06N3/08G06K9/62
CPCG06F21/6245G06N3/084G06N3/044G06N3/045G06F18/24
Inventor 陶陶柏建树郑啸刘恒王爱国
Owner ANHUI UNIVERSITY OF TECHNOLOGY
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

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

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com