异构数据下具有拜占庭鲁棒性的轻量级安全联邦学习方法
By constructing a global model and a system model, and utilizing dual-server collaborative gradient aggregation, the problem of difficult identification of malicious gradients under heterogeneous data is solved. This achieves robustness and privacy protection of secure federated learning in Non-IID scenarios, improving the accuracy and security of the model.
CN118036779BActive Publication Date: 2026-07-17UNIV OF ELECTRONICS SCI & TECH OF CHINA
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2024-02-06
- Publication Date
- 2026-07-17
Smart Images
Figure CN118036779B_ABST
Abstract
本发明公开了一种异构数据下具有拜占庭鲁棒性的轻量级安全联邦学习方法，包括以下步骤：S1，构建全局模型的目标函数；S2，建立全局模型的安全协同计算的系统模型；S3，拟定多个用于系统模型的安全计算协议；S4，在安全计算协议下，利用系统实现用户安全计算本地梯度和双服务器协同计算梯度聚合，从而完成全局模型的协同训练。通过上述设计，本发明利用用户本地模型与全局模型的二范数作为本地目标函数的惩罚项，以约束本地梯度过于发散，提高在Non‑IID场景下恶意梯度和真实梯度的区分度。
Need to check novelty before this filing date? Find Prior Art