A federated learning method and device, electronic equipment and storage medium

CN114996772BActive Publication Date: 2026-05-29SHANGHAI FUSHU TECH CO LTD
2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHANGHAI FUSHU TECH CO LTD
Filing Date
2022-06-17
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

The existing federated learning process has low security, especially since unlabeled devices can easily crack the sample labels on labeled devices by using gradient values ​​and predicted labels.

Method used

Noise values ​​are added to the gradient random values. The noise values ​​are generated by homomorphic encryption and local differential privacy algorithms. The noisy gradient random values ​​are then sent to the unlabeled device for federated learning.

Benefits of technology

This improves the security of federated learning, making it difficult for unlabeled devices to crack sample labels on labeled devices, thus protecting data privacy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114996772B_ABST
    Figure CN114996772B_ABST
Patent Text Reader

Abstract

The application provides a federated learning method and device, electronic equipment and storage medium, which are used for improving the low security of the current federated learning process. The method comprises the following steps: receiving a ciphertext gradient random value sent by a non-labeled party device, the non-labeled party device storing sample data, and the ciphertext gradient random value being obtained by homomorphic encryption of a loss value corresponding to the sample data and adding a random number; homomorphic decryption is performed on the ciphertext gradient random value to obtain a gradient random value; a noise value is added to the gradient random value to obtain a noisy gradient random value; and the noisy gradient random value is sent to the non-labeled party device, so that the non-labeled party device performs federated learning on a local model according to the noisy gradient random value.
Need to check novelty before this filing date? Find Prior Art