VERTICAL FEDERATED LEARNING METHODS, APPARATUSES, SYSTEM AND DEVICE, AND STORAGE MEDIUM

WIPO Patent Application WO/2023/174018

A1

Provided in the present disclosure are vertical federated learning methods, apparatuses, system and device, and a storage medium. A method comprises: a first data party calculates a noise matrix on the basis of a mask matrix, determines the product of a residual vector and the noise matrix as a noise-added residual vector, and sends the noise-added residual vector to a second data party. The second data party calculates a gradient vector on the basis of the noise-added residual vector to update model parameters. In the present disclosure, the first data party calculates the noise matrix and encrypts the residual vector on the basis of the noise matrix for the second data party, thus ensuring that the residual vector calculated by the first data party will not be acquired by the second data party and achieving the purpose of protecting the privacy of labels in samples of the first data party. In addition, the computation overhead is relatively low due to the mode of encrypting a residual vector by means of a noise matrix, so that the present disclosure improves the efficiency of vertical federated learning while ensuring data privacy.

HE PEIXUAN (CN)
ZHANG YAO (CN)
LIU YANG (CN)
WU YE (CN)

PCT/CN2023/077525

September 21, 2023

February 22, 2023

Click for automatic bibliography generation  

BEIJING BYTEDANCE NETWORK TECH CO LTD (CN)

G06F21/60

CN114611128A	2022-06-10
CN112132293A	2020-12-25
CN112182594A	2021-01-05
CN112149171A	2020-12-29
CN112906912A	2021-06-04
CN112199702A	2021-01-08
US20210374605A1	2021-12-02

ZHIFAN & PARTNERS (CN)

View/Download PDF      PDF Help

Previous Patent: ONLINE AUCTION METHOD, DEVICE, APPARATUS, AND MEDIUM

Next Patent: IMPELLER STRUCTURE, HEART PUMP, AND HEART ASSIST SYSTEM