Paper 2023/1544
Arithmetic PCA for Encrypted Data
Abstract
Reducing the size of large dimensional data is a critical task in machine learning (ML) that often involves using principal component analysis (PCA). In privacy-preserving ML, data confidentiality is of utmost importance, and reducing data size is a crucial way to cut overall costs. This work focuses on minimizing the number of normalization processes in the PCA algorithm, which is a costly procedure in encrypted PCA. By modifying Krasulina's algorithm, non-polynomial operations were eliminated, except for a single delayed normalization at the end. Our PCA algorithm demonstrated similar performance to conventional PCA algorithms in face recognition applications. We also implemented it using the CKKS (Cheon-Kim-Kim-Song) homomorphic encryption scheme and obtained the first 6 principal components of a 128$\times$128 real matrix in 7.85 minutes using 8 threads.
Metadata
- Available format(s)
-
PDF
- Category
- Cryptographic protocols
- Publication info
- Preprint.
- Keywords
- Principal Component AnalysisHomomorphic EncryptionPrivacy-Preserving Machine Learning
- Contact author(s)
-
jhcheon @ snu ac kr
sixtail528 @ snu ac kr
newstar0329 @ snu ac kr
duhyeong kim @ intel com
dahoon46 @ snu ac kr
jhyunp @ snu ac kr - History
- 2023-10-09: approved
- 2023-10-09: received
- See all versions
- Short URL
- https://ia.cr/2023/1544
- License
-
CC BY
BibTeX
@misc{cryptoeprint:2023/1544,
author = {Jung Hee Cheon and Hyeongmin Choe and Saebyul Jung and Duhyeong Kim and Dah Hoon Lee and Jai Hyun Park},
title = {Arithmetic PCA for Encrypted Data},
howpublished = {Cryptology ePrint Archive, Paper 2023/1544},
year = {2023},
note = {\url{https://eprint.iacr.org/2023/1544}},
url = {https://eprint.iacr.org/2023/1544}
}
