Federal learning model updating method and device

Abstract

The invention provides a federal learning model updating method and device, and the method comprises the steps: transmitting a first encryption gradient obtained by encrypting a first gradient through employing a first public key to an active terminal and a passive terminal of each first group, sending a second encryption gradient obtained by encrypting the first gradient by using a second public key to other terminals of a second group where the current terminal is located; receiving third encryption gradients respectively sent by the active terminal, the passive terminal and other terminals of each first group; determining a first cipher text gradient corresponding to each second public key according to a third encryption gradient obtained by encrypting the same public key; sending the first cipher text gradient to an active terminal containing a private key for decrypting the first cipher text gradient; and receiving a first decryption gradient fed back by each active terminal, and updating parameters of the model according to each first decryption gradient and a first noise value. According to the federal learning model updating method provided by the invention, the stability of model training is improved on the premise of ensuring the data privacy.

Description

technical field [0001] The invention relates to federated learning technology, in particular to a method and device for updating a federated learning model. Background technique [0002] In order to solve the problems of data islands and data privacy security, the current mainstream method is to use federated learning to jointly train different data to obtain better models to solve practical problems. According to the distribution of data, federated learning can be divided into horizontal federated learning, vertical federated learning, and transfer learning. Among them, vertical federated learning is widely used. For example, in financial scenarios, financial institutions such as banks contain credit labels, while e-commerce platforms have user consumption data, and users of both parties overlap. Banks can use e-commerce data to predict credit risks, but the two parties cannot share data. Vertical federated learning can be used to solve such problems. [0003] At present...

AI Technical Summary

Problems solved by technology

[0005] The invention provides a method and device for updating a federated learning model to solve the problem of poor training stability of the model

Images