基于秘密共享的轻量级隐私保护ViT推理框架
Light weighted privacy protection ViT inference framework based on secret sharing
马敏 1付钰 2黄凯 3贾潇风4
作者信息
- 1. 海军工程大学信息安全系,湖北 武汉 430033;湖北开放大学软件工程学院,湖北 武汉 430074
- 2. 海军工程大学信息安全系,湖北 武汉 430033
- 3. 国防大学联合作战学院,河北 石家庄 050084
- 4. 浙江工商大学计算机科学与技术学院,浙江 杭州 310018
- 折叠
摘要
针对广泛应用于图像处理的ViT推理框架存在泄露用户隐私数据的风险,而已有隐私保护推理框架存在计算效率较低、在线通信量较大等问题,提出了一种高效隐私保护推理框架SViT.该框架由2个边缘服务器协作执行基于秘密共享设计的安全计算协议SSoftmax、SLayerNorm、SGeLU,在保持ViT-B/16原始框架结构的情况下,解决了隐私保护框架推理开销大的问题.理论分析与实验表明,相比CrypTen,SViT在计算效率和在线通信开销方面分别提升了2～6倍和4～14倍.
Abstract
The ViT(vision transformer)inference framework,which was widely used in image processing,was found to have a risk of leaking user privacy data.However,existing privacy protection inference frameworks had problems such as low computational efficiency and high online communication volume.To address this issue,a highly efficient privacy protection inference framework SViT was proposed.Two edge servers collaborated to execute secure computing proto-cols based on secret sharing design,such as SSoftmax,SLayerNorm,SGeLU,etc.While maintaining the original frame-work structure of ViT-B/16,the problem of large inference overhead in privacy protection framework was solved.Theo-retical analysis and experiments show that compared to Crypton,SViT has improved computational efficiency by 2～6 times and online communication overhead by 4～14 times,respectively.
关键词
隐私保护/秘密共享/图像分类/安全计算协议Key words
privacy protection/secret sharing/image classification/secure computing protocol引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
维普
万方数据