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

VHE-based privacy protection neural network training and predicting method

A technology of neural network and prediction method, applied in the field of vector homomorphic computing, can solve problems such as data security cannot be guaranteed

Inactive Publication Date: 2018-11-09
UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF2 Cites 13 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a VHE-based privacy protection neural network training and prediction method, which solves the technical problem that data security cannot be guaranteed when using BP neural network for training and prediction.

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
  • VHE-based privacy protection neural network training and predicting method
  • VHE-based privacy protection neural network training and predicting method
  • VHE-based privacy protection neural network training and predicting method

Examples

Experimental program
Comparison scheme
Effect test

specific Embodiment 1

[0109] In this embodiment, the number of hidden layers in the BP neural network is 1, then the learning rate is set to α, and two weight matrices w are initialized 1 and w 2 ;

[0110] w 1 Represents the weight matrix between the input layer and the hidden layer, with a size of (the input vector dimension of the data set D plus 1, the number of neurons in the hidden layer);

[0111] w 2 Indicates the weight matrix between the hidden layer and the output layer, the size is (the number of neurons in the hidden layer, the number of neurons in the output layer)

[0112] Simultaneously initialize 2 threshold vectors b 1 and b 2 , b 1 Represents the threshold vector of the hidden layer, b 2 Represents the threshold vector of the output layer, all initialized to all zero vectors;

[0113] Define the activation function f of the hidden layer 1 and the activation function f of the output layer 2 ;

[0114] The input to the hidden layer is:

[0115] The output of the hidde...

specific Embodiment 2

[0141] Based on the specific embodiment 1, this embodiment uses data for further description.

[0142] The input of data set D =

[0143](8 307 130 3504 12 70 1; 8 350 165 3693 11.5 70 1; 8 318 150 3436 11 701; 8 304 150 3433 12 70 1; 8 302 140 3449 10.5 70 1; 829 198 4341 10 70 1; 8454 22020 4354 9 70 1; 8 440 215 4312 8.5 70 1; 8 455 225 4425 10 70 1; 8 390 1903850 8.5 70 1; 8 383 170 3563 10 70 1; 8 340 160 3609 80 1; ;8 455 225 3086 10 70 1; 4 113 95 2372 15 70 3; 6 198 95 2833 15.5 70 1; 6199 97 2774 15.5 70 1; 461835 20.5 70 2; 4 110 87 2672 17.5 70 2; 4 107 90 2430 14.5 70 2; 4 104 95 237517.5 70 2; 4 121 113 2234 12.5 70 2; 6 19990 2648 15 70 1; 8 360 215 4615 14 701 ;8 307 200 4376 15 70 1; 8 318 210 4382 13.5 70 1; 8 304 193 4732 18.5 70 1; 497 88 2130 14.5 71 3; 1002634 13 71 1; 6 225 105 3439 15.5 71 1; 6 250 100 3329 15.5 71 1; 6 250 88 330215.5 71 1; 6 232 100 3288 15.5 71 1; 8 350 165 4209 12 71 1 1; 8 400 175 4464 11.571 1 1 1 ;8 351 153 4154 13.5 71 1; 8 3...

specific Embodiment 3

[0155] Based on specific embodiment 1, this embodiment uses data for further description.

[0156] The input of data set D =

[0157] ([[-3, -2.7, -2.4, -2.1, -1.8, -1.5, -1.2, -0.9, -0.6, -0.3, 0, 0.3, 0.6, 0.9, 1.2, 1.5, 1.8], [- 2, -1.8, -1.6, -1.4, -1.2, -1, -0.8, -0.6, -0.4, -0.2, -2.2204, 0.2, 0.4, 0.6, 0.8, 1, 1.2])

[0158] The output of dataset D =

[0159] ([0.6589, 0.2206, -0.1635, -0.4712, -0.6858, -0.7975, -0.8040, -0.7113, -0.5326, -0.2875, 0, 0.3035, 0.5966, 0.8553, 1.0600, 1.1975, 1.2618])

[0160] Among them, the data is first uniformly enlarged by 10000 times and truncated to an integer, and then encrypted with VHE; there are 2 neurons in the input layer, 3 neurons in the hidden layer, and 1 neuron in the output layer; the activation function of the hidden layer is a nonlinear function y= tanh x, the data is reduced by 400 times before passing the tanhx function, and then enlarged by 400 times after passing the tanh x function; the output layer excitation ...

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 VHE-based privacy protection neural network training and predicting method. The method comprises the following steps of: encrypting a data set D by utilizing a VHE homomorphic encryption algorithm so as to obtain an encrypted data set D', wherein the encrypted data set D' comprises a training data set D'1 and a test data set D'2; carrying out BP neural network batch gradient training on the training data set D'1 so as to obtain a trained BP neural network; and predicting the test data set D'2 by utilizing the trained BP neural network. According to the method, an encryption algorithm and a BP neural network method are combined to realize the training and prediction of BP neural networks under a cyphertext domain, namely, BP neural network training and prediction also can be carried out on data under encryption protection when clouds are not credible, so that real computing outsourcing is realized.

Description

technical field [0001] The invention relates to the field of vector homomorphic calculation, in particular to a training and prediction method of a VHE-based privacy protection neural network. Background technique [0002] Today, big data has become the trend of future social and economic development, and has great application potential in various fields of human society. Huge amounts of data from sensors, social networks, businesses, the internet, etc. are collected, shared and analyzed. Due to the widespread use of cloud computing, users are now outsourcing not only their data but also data mining tasks to the cloud. However, the cloud is very likely to be insecure, and the cloud may jeopardize the user's privacy during data mining. The mined data can contain sensitive information such as personal identities, medical records and even financial information. [0003] The BP neural network is a multi-layer feed-forward network. The main work is divided into two stages: the...

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): G06N3/08G06F21/60G06F21/62
CPCG06N3/084G06F21/602G06F21/6245
Inventor 杨浩淼张有何伟超梁绍鹏李洪伟任彦之
Owner UNIV OF ELECTRONICS SCI & TECH OF CHINA
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