Unlock instant, AI-driven research and patent intelligence for your innovation.

Model training method and system based on data encryption

A model training and data encryption technology, applied in the field of data processing, can solve problems such as inability to flexibly satisfy users, inability to filter data set features, etc., achieve strong data selection flexibility, save training time, and improve accuracy.

Active Publication Date: 2021-11-09
UNIV OF JINAN
View PDF5 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in the past machine learning algorithms combined with data privacy protection, the complete data set uploaded to the cloud is used for model training, and it is impossible to filter a certain feature of the data set and flexibly meet the needs of users.

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
  • Model training method and system based on data encryption
  • Model training method and system based on data encryption
  • Model training method and system based on data encryption

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0057] Embodiment 1 provides a model training system based on data encryption, which includes:

[0058] The first encryption module is used to encrypt the original data using the public key to obtain the ciphertext c;

[0059] The second encryption module is used to generate a random negative number R and the opposite number R of the random negative number 1 , where R is used as the plaintext share required for training, and the opposite number R of the random negative number is encrypted using the public key 1 Get the ciphertext share c 2 ;

[0060] Homomorphic operation module, used to combine c and c 2 Generate a secret shared ciphertext share c through homomorphic operation 3 ;

[0061] Decryption module for decrypting the ciphertext share c using a private key 3 , get the plaintext share m 3 ;

[0062] Interactive operation module for two parts of plaintext shares R and m 3 , with the help of Beaver triples for interactive training, two model shares are obtained;...

Embodiment 2

[0093] Embodiment 2 provides a model training system based on data encryption, which includes:

[0094] The first encryption module is used to encrypt the original data using the public key to obtain the ciphertext c;

[0095] The second encryption module is used to generate a random negative number and the opposite number of the random negative number, and use the public key to encrypt the opposite number of the random negative number to obtain the ciphertext share c 2 ;

[0096] Homomorphic operation module, used to combine c and c 2Generate a secret shared ciphertext share c through homomorphic operation 3 ;

[0097] Decryption module for decrypting the ciphertext share c using a private key 3 , get the plaintext share m 3 ;

[0098] Interactive computing module for using random negative numbers and plaintext shares m 3 , with the help of Beaver triples for interactive training, two model shares are obtained;

[0099] The combination module is used to add the shares...

Embodiment 3

[0129] like figure 1 As shown, in order to solve the problem of data storage and training security and efficiency on the third-party server and whether the obtained model can meet the needs of users, this embodiment 3 proposes a method that enhances data privacy protection and improves model training at the same time. Model approach for flexibility and efficiency.

[0130] In the method of Embodiment 3, the data owner extracts keywords, builds a lookup table, a storage table, encrypts data with a public key, and uploads the lookup table, storage table, and ciphertext to the server S1. The queryer extracts keywords according to his or her own needs, builds a query trapdoor, and sends the trapdoor to the server. After receiving the query trapdoor, the server finds the corresponding ciphertext and screens out the data required by the queryer; then generates a random number locally, encrypts it with the public key, performs homomorphic operation with the required data to gen...

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 model training method and system based on data encryption, and belongs to the technical field of data processing, and the method comprises the steps: encrypting original data through employing a public key, and obtaining a ciphertext c; generating a random negative number R and an opposite number R1 of the random negative number, and encrypting the opposite number R1 of the random negative number by using a public key to obtain a ciphertext share c2; c and c2 are subjected to homomorphic operation to generate a secretly shared ciphertext share c3; decrypting the ciphertext share c3 by using a private key to obtain a plaintext share m3; performing interactive training by using a random negative number and a plaintext share m3 and by means of a Beaver triple to obtain two model shares; and adding the two obtained model shares to obtain a final required model. According to the method, data can be searched for training in a targeted manner according to self requirements, and high data selection flexibility is achieved; in the data storage and training, the data exists in a ciphertext or plaintext share, so that the security of the data is ensured; before the model training, the data is converted from the homomorphic ciphertext to the plaintext share, so that the high efficiency and accuracy of the model training are ensured.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a model training method and system based on data encryption. Background technique [0002] In traditional machine learning, users need to directly upload plaintext data to the server, but many servers on the Internet that contain a large amount of data have serious security vulnerabilities, and some cloud service providers (such as Google, Facebook) will automatically mine user privacy data, This leads to the leakage of user data privacy information. Therefore, the data is directly stored in the cloud server in plain text, which will endanger the security of user information. [0003] Using homomorphic encryption technology, the data can be uploaded to the server in the form of ciphertext, and the calculation of real numbers can be optimized, and then the least square method can be used to approximate the sigmod function to improve accuracy and efficiency (that is, reduce...

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): H04L9/30H04L9/00H04L29/06G06F21/60
CPCH04L9/30H04L9/008H04L63/0435G06F21/602
Inventor 赵川贾海鑫荆山刘伟
Owner UNIV OF JINAN