Supercharge Your Innovation With Domain-Expert AI Agents!

Cross-organization distributed deep learning method based on homomorphic encryption

A homomorphic encryption and deep learning technology, applied in neural learning methods, biological neural network models, neural architectures, etc., can solve the problems of accelerating a single homomorphic encryption operation or compressing the space occupied by ciphertext, and unable to eliminate bottlenecks

Pending Publication Date: 2022-08-02
SHANGHAI UNIVERSITY OF ELECTRIC POWER
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Since speeding up a single homomorphic encryption operation or compressing the ciphertext footprint does not eliminate the bottleneck of adopting homomorphic encryption in distributed deep learning

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
  • Cross-organization distributed deep learning method based on homomorphic encryption
  • Cross-organization distributed deep learning method based on homomorphic encryption
  • Cross-organization distributed deep learning method based on homomorphic encryption

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0040] refer to Figure 1-6 The invention discloses a cross-organization distributed deep learning method based on homomorphic encryption. The core of the method is to establish a distributed deep learning system architecture, wherein the homomorphic encryption is implemented as a pluggable module on the client, and the aggregator is the server that coordinates clients and aggregates their cryptographic gradients. Before training starts, the aggregator randomly selects a client as the leader to generate a homomorphic cryptographic key pair and synchronize it to all other clients. The leader also initializes the deep learning model and sends the model weights to all other clients. After receiving the homomorphic encryption key pair and initial weights, the client starts training. In one iteration, each client computes a local gradient update, encrypts it with the public key, and transmits the result to the server, which collects the gradients from all clients, then, it adds 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 discloses a cross-organization distributed deep learning method based on homomorphic encryption, and the method comprises the following steps: randomly selecting a client to generate a homomorphic encryption key pair, and synchronizing the homomorphic encryption key pair to all other clients; the selected leader client sends the initialized model to the other clients; the client performs deep learning iteration and local gradient calculation; and the server collects the processed gradient data from the client and updates the model by using the data. According to the method and the framework on which the method depends, serious calculation and communication overhead caused by homomorphic encryption can be greatly reduced, and meanwhile it is guaranteed that the model prediction accuracy is not damaged.

Description

technical field [0001] The invention relates to the technical field of distributed deep learning, in particular to a cross-organizational distributed deep learning method based on homomorphic encryption. Background technique [0002] Deep learning has come a long way in recent years. Researchers and engineers have applied deep learning techniques to many fields including computer vision, natural language processing, speech recognition, and more. In order to obtain higher prediction accuracy and support smarter tasks, more complex neural networks need to be trained. However, the input data required to train a large model grows exponentially with the model parameters, and training large deep neural networks on large-scale data has exceeded the computing and storage capabilities of a single machine; Doing centralized training can lead to serious privacy concerns and may even be prohibited by regulations. Therefore, there is a need to distribute the training workload among mu...

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): G06F21/60G06N3/04G06N3/08
CPCG06F21/602G06N3/04G06N3/084
Inventor 杜海舟冯晓杰
Owner SHANGHAI UNIVERSITY OF ELECTRIC POWER
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