Federated deep learning method capable of adaptively protecting privacy

A privacy protection and deep learning technology, applied in machine learning, digital data protection, instruments, etc., to achieve the effect of increasing accuracy, increasing privacy, and improving accuracy

Active Publication Date: 2019-11-12
UNIV OF ELECTRONIC SCI & TECH OF CHINA
View PDF6 Cites 58 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the differential privacy mechanism ne

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
  • Federated deep learning method capable of adaptively protecting privacy
  • Federated deep learning method capable of adaptively protecting privacy
  • Federated deep learning method capable of adaptively protecting privacy

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0025] 1. The system model of the invention is as figure 1 Shown:

[0026] 2. System initialization, including the following steps:

[0027] 1) Participant U g A learning model that is more accurate and does not lead to local overfitting needs to be jointly obtained, and the participant U g Negotiate a deep learning network with the cloud server in advance, such as convolutional neural network CNN, recurrent neural network RNN, etc.;

[0028] 2) The server uses some public data for training according to the user data type, and obtains an initialized deep learning model parameter w global ;

[0029] 3. The participant initializes the local model, which is characterized in that it includes the following steps:

[0030] 1) The server broadcasts its own deep learning model parameter w global ;

[0031] 2) Each participant U g Download the initialized model parameters w global , and update its own local learning model w local ;

[0032] 4. The participant preprocesses th...

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 provides a federated deep learning method capable of adaptively protecting privacy, which is used for protecting original data of a user from being known by a curious server in federateddeep learning and protecting parameters of a learning model from leaking information of the original data of the user. The method comprises: each participant negotiating a network framework with thecloud server in advance, then the cloud server obtaining an initialized model, and the cloud server broadcasting model parameters to each participant; the participants downloading the initialized model parameters and updating local models of the participants, then performing training in combination with a local data set, performing different privacy protection operations on different data featuresbased on different contribution degrees of data attributes to model output, and the participants sending local gradients obtained through respective training to the cloud server; and finally, the cloud server collecting the gradient information of each participant and then updating the model of the participant to carry out subsequent training. On the premise of meeting privacy protection, the accuracy of the learning model is greatly improved.

Description

technical field [0001] The present invention relates to artificial intelligence technology. Background technique [0002] Traditional centralized deep learning requires user data to be concentrated in a data center, and users lose control over their own data, and this part of data may also be abused by data users or speculate on more user privacy information. Federated Deep Learning proposed by Google can solve the privacy, location, usage rights and other issues of user data. [0003] Federated Deep Learning allows multiple participants to jointly learn a common model without disclosing their own data sets. This requires participants to train with their own local data sets to obtain a local model, and share their training gradients with other participants; the cloud server or a user aggregates the training gradients from each participant, and then obtains a "common" model, which also prevents local overfitting of the user's local model. [0004] Differential Privacy Mech...

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): G06F21/62G06N20/00
CPCG06F21/6245G06N20/00
Inventor 李洪伟刘小源徐国文刘森龚丽姜文博成艺任彦之李双
Owner UNIV OF ELECTRONIC SCI & TECH OF CHINA
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