Federated learning private data processing method, device and system

Abstract

The invention discloses a federated learning private data processing method. The private data processing method comprises the following steps: receiving aggregation model parameters sent by the coordination device or other participation devices; obtaining training data, generating an updating model, and outputting updating model parameters of the updating model; using an encryption algorithm to perform cryptographic calculation on the updated model parameters to generate ciphertext data, in the process of cryptographic calculation, splitting a first operand into a plurality of second operands,the bit number of the first operand being greater than the bit number of the second operands; and sending the ciphertext data to the coordinator or other participants to obtain a global model. According to the invention, the first operand in ciphertext calculation is split into the plurality of second operands, so that the requirement of a machine instruction is met, and the execution of a federated learning ciphertext calculation task can be realized on a universal processor.

Description

technical field [0001] The present invention relates to the field of private data encryption processing, and in particular to a federated learning private data processing method, device and system. Background technique [0002] In federated learning applications, participating devices need to send updated model parameters (for example, neural network model weights, or gradient information) to the coordinating device, so that the updated model parameters will be informed by the coordinator. In the scenario where the reliability of the coordination equipment cannot be guaranteed, the private data information of the participating equipment may be leaked. In order to ensure that the private data information of the participating devices is not leaked to the coordinating device, the participating devices can encrypt the private data information through encryption algorithms such as Paillier or RSA. [0003] Traditional machine learning generally uses 32-bit basic operations, whic...

AI Technical Summary

Problems solved by technology

[0003] Traditional machine learning generally uses 32-bit basic operations, which are generally directly supported by machine instructions, while data operations encrypted by Paillier or RSA algorithms used in federated learning are 1024-bit or 2048-bit or even Longer large integer operations, and current machine instructions do not support such large integer operations

Images