福建电脑2025，Vol.41Issue(11)：1-7,7.DOI:10.16707/j.cnki.fjpc.2025.11.001
循环神经网络安全推理的同态加密技术
A Survey on the Development of Secure Recurrent Neural Networks Based on Homomorphic Encryption
石志都1
作者信息
- 1. 福州大学计算机与大数据学院 福州 350108
- 折叠
摘要
Abstract
This paper provides an overview of the research progress in secure recurrent neural networks based on homomorphic encryption from 2019 to 2024.The research focuses on four technical directions:activation function approximation,architecture reconstruction,computation acceleration,and resource optimization.The optimal solution achieves a 360-fold increase in throughput with only a 0.1%-5%loss in accuracy.However,existing methods still face challenges such as high computational complexity and limited sequence processing.In the future,it is necessary to break through the limitations of computational depth,combine hardware acceleration with algorithm optimization,and promote practical applications in fields such as healthcare and finance.关键词
同态加密/循环神经网络/隐私保护Key words
Homomorphic Encryption/Recurrent Neural Network/Privacy-Preserving分类
计算机与自动化引用本文复制引用
石志都..循环神经网络安全推理的同态加密技术[J].福建电脑,2025,41(11):1-7,7.
