A Privacy Computing Method Based on Heterogeneous Neural Network Model

A neural network model and computing method technology, applied in biological neural network models, neural learning methods, digital data protection, etc., can solve problems such as reducing the accuracy of neural network models, improve training efficiency, promote flow, and expand applications. range effect

Active Publication Date: 2022-02-22
浙江数秦科技有限公司
View PDF4 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Its technical solution improves the privacy protection of input data when the neural network model is executed, but reduces the accuracy of the neural network model

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
  • A Privacy Computing Method Based on Heterogeneous Neural Network Model
  • A Privacy Computing Method Based on Heterogeneous Neural Network Model
  • A Privacy Computing Method Based on Heterogeneous Neural Network Model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0031] A privacy calculation method based on a heterogeneous neural network model, the heterogeneous neural network model is a neural network model comprising at least one multiplication neuron 15, please refer to the attached figure 1 , the privacy calculation method includes: step A01) establishing a service node and an execution node, several participants participating in the privacy calculation provide an objective function or sample data to the service node, and the service node establishes and trains a heterogeneous neural network model according to the objective function or sample data . Step A02) The service node constructs several sub-models, which correspond to the neurons of the first layer one by one, the input of the sub-model is the neuron of the 0th layer connected to the corresponding neuron of the first layer, and the output of the sub-model is the first layer The input number of neurons, the sub-model corresponding to multiplication neuron 15 is recorded as t...

Embodiment 2

[0053] A privacy calculation method based on a heterogeneous neural network model. On the basis of the first embodiment, this embodiment provides an improved solution method for the heterogeneous neural network model, so that the main model can obtain a confidentiality effect. Please refer to the attached Figure 11 , on the basis of Embodiment 1, the embodiment also includes the following steps: Step F01) The service node generates a confusion coefficient for each multiplication neuron 15, and divides the weight coefficients of the connections involved in the multiplication neuron 15 in the main model by the corresponding Then the main model is sent to the execution node; Step F02) The input neuron of the multiplication sub-model is set with an adjustment coefficient, and the privacy number of the participant is multiplied by the adjustment coefficient and then split into several multipliers ; Step F03) The service node generates the adjustment coefficient of each input neuro...

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 present invention relates to the technical field of privacy computing, in particular to a privacy computing method based on a heterogeneous neural network model, including: establishing a service node and an execution node, the participant provides an objective function or sample data to the service node; the service node establishes and trains a different Construct a neural network model; build several sub-models and the main model; send the number of the sub-models to several participants, and send the main model to the execution node; the participants split their privacy numbers into the sum of several addends and several The product of multiple multipliers; the participants use their assigned addends and the output of the multiplier calculation sub-model as the intermediate value, and send the intermediate value to the execution node; the execution node multiplies the intermediate value of the multiplication sub-model, and the remaining sub-models The intermediate values ​​of are added together to obtain the output of the sub-model, which is substituted into the main model to obtain the privacy calculation result. The substantive effect of the present invention is to improve the execution efficiency of privacy calculation, realize data availability and invisible, and promote the flow of data elements.

Description

technical field [0001] The invention relates to the technical field of privacy computing, in particular to a privacy computing method based on a heterogeneous neural network model. Background technique [0002] Privacy-preserving computation (Privacy-preserving computation) refers to a series of information technologies that analyze and compute data on the premise that the data provider does not disclose the original data, ensuring that the data is "available and invisible" in the process of circulation and fusion. Privacy computing is an important way to solve the problem of serious privacy leakage risks while exerting the value of data. The current privacy computing technologies mainly include secure multi-party computing technology based on obfuscated circuits and inadvertent transmission, and homomorphic encryption computing based on homomorphic encryption technology. However, when the implementation of the confusion circuit is complex or the value span is large, the es...

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 Patents(China)
IPC IPC(8): G06F21/60G06N3/08
CPCG06F21/60G06N3/08
Inventor 张金琳高航俞学劢
Owner 浙江数秦科技有限公司
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