Automorphism Architecture for Bootstrapping Homomorphic Encryption

Citations

WEB OF SCIENCE

0
Citations

SCOPUS

0

초록

In Fully Homomorphic Encryption (FHE), Automorphism used to map very long vectors to new positions can exhibit varying patterns depending on the rotation index. However, increasing the polynomial degree introduces significant overhead in terms of both execution time and hardware resources. Furthermore, the data mapping process for each moduli generated by the Residue Number System (RNS) becomes dominant in terms of execution time. Therefore, RNS-based homomorphic encryption hardware architectures need to support processing data mapping in parallel while communicating with on-chip memory, and all rotation indices for bootstrapping. The proposed novel Automorphism architecture makes efficient use of FPGA resources and supports 8 parallel, so that the Batch Data becomes 128, and uses Xilinx's Built-In FIFO for intermediate storage. It also supports all rotation indices calculated by Cyclotomic degree (M=2*N). © 2024 IEEE.

키워드

AutomorphismBootstrappingFully Homomorphic Encryption (FHE)Key-SwitchingNumber Theoretic Transform (NTT)Residue Number System (RNS)Rotation
제목
Automorphism Architecture for Bootstrapping Homomorphic Encryption
저자
Lee, HanyoungLee, Hanho
DOI
10.1109/ISOCC62682.2024.10762582
발행일
2024
유형
Proceedings Paper
저널명
Proceedings - International SoC Design Conference 2024, ISOCC 2024
페이지
159 ~ 160