基于同态密文转换的隐私保护卷积神经网络推理方案
Privacy-preserving convolutional neural network inference scheme based on homomorphic ciphertext transformation
李瑞琪 1易琴 1黄艺璇 2贾春福3
作者信息
- 1. 中国民航大学安全科学与工程学院,天津 300300
- 2. 南开大学网络空间安全学院,天津 300350
- 3. 南开大学网络空间安全学院,天津 300350;天津市网络与数据安全技术重点实验室,天津 300350
- 折叠
摘要
为了解决现有隐私保护卷积神经网络交互频繁、推理准确率稍低等问题,基于同态密文转换框架,提出了一种同态友好型的非交互式隐私保护卷积神经网络推理方案.利用Pegasus同态密文转换框架,在卷积层中利用CKKS(Cheon-Kim-Kim-Song)密文进行并行化的卷积运算;在激活层和池化层中利用LWE密文和LUT(look-up table)技术实现激活函数、最大池化和全局池化的计算;利用Pegasus框架提供的密文转换技术,实现不同形式的同态密文之间的转换.理论分析和实验结果表明,所提方案能够保证数据安全,并且具有较高的推理准确率和较低的计算和通信开销.
Abstract
To solve the problems of frequent interaction and low prediction accuracy of existing privacy-protected convo-lutional neural networks,a homomorphic friendly non-interactive privacy-protected convolutional neural network infer-ence scheme was proposed via homomorphic ciphertext transformation.Utilizing the Pegasus framework,CKKS(Cheon-Kim-Kim-Song)ciphertext was used to parallelize convolution operations in convolution layer.In the activation layer and pooling layer,LWE ciphertext and LUT(look-up table)technology were used to calculate the activation func-tion,maximum pooling and global pooling.Using the ciphertext conversion technology provided by the Pegasus frame-work,the conversion between different forms of homomorphic ciphertext is realized.Theoretical analysis and experimen-tal results show that the proposed scheme can ensure data security,and has higher inference accuracy and lower calcula-tion and communication overhead.
关键词
隐私保护/卷积神经网络/同态加密/密文转换Key words
privacy-preserving/convolutional neural network/homomorphic encryption/ciphertext transformation引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据