An eDRAM-Based In-Situ-Computing Processor for Homomorphic Encryption Evaluation on the Edge

This work presents an in-situ-computing accelerator for homomorphic encryption (HE) evaluation on the edge, with eight eDRAM-based in-situ HE processing cores integrated with RISC-V CPU and HE-specialized NoC. The key features are: (1) the architecture-level fusion with customized ISA chains HE operators, avoiding excessive intermediate ciphertext movements; (2) the circuit-level fusion of eDRAM bitcells and dynamic digital datapath enables energy-efficient in-memory HE evaluation with high storage density; (3) intra-/inter-core hybrid automorphism network greatly reduces communication overheads. The silicon prototype in 28 nm CMOS achieves $2.33 \mathrm{Mb} / \mathrm{mm}^{2} \mathrm{HE}$ storage density, with peak energy efficiency of 329 nJ/NTT at 4096-points-19-bit and 15.1 classification per second with a linear support vector machine in CKKS-RNS scheme, exceeding previously reported ASIC works and CPUs.

Authors

Institutions

Publication Details

Published
2024-09-09
DOI
https://doi.org/10.1109/esserc62670.2024.10719576
Citations
2
Primary Topic
Advanced Data Storage Technologies
Type
article
Field-Weighted Citation Impact
0.70
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

An eDRAM-Based In-Situ-Computing Processor for Homomorphic Encryption Evaluation on the Edge

Luchang Lei, Huazhong Yang, Yongpan Liu, Xinchen Zou et al.
2 citations
Advanced Data Storage Technologies
0.70
article

An eDRAM-Based In-Situ-Computing Processor for Homomorphic Encryption Evaluation on the Edge

Luchang Lei, Huazhong Yang, Yongpan Liu, Xinchen Zou, Song Bian, Yongqing Zhu, Zhenyu Guan, Hongyang Jia, Zhou Zhang, Yifan He
article en
2 citations

Abstract

This work presents an in-situ-computing accelerator for homomorphic encryption (HE) evaluation on the edge, with eight eDRAM-based in-situ HE processing cores integrated with RISC-V CPU and HE-specialized NoC. The key features are: (1) the architecture-level fusion with customized ISA chains HE operators, avoiding excessive intermediate ciphertext movements; (2) the circuit-level fusion of eDRAM bitcells and dynamic digital datapath enables energy-efficient in-memory HE evaluation with high storage density; (3) intra-/inter-core hybrid automorphism network greatly reduces communication overheads. The silicon prototype in 28 nm CMOS achieves $2.33 \mathrm{Mb} / \mathrm{mm}^{2} \mathrm{HE}$ storage density, with peak energy efficiency of 329 nJ/NTT at 4096-points-19-bit and 15.1 classification per second with a linear support vector machine in CKKS-RNS scheme, exceeding previously reported ASIC works and CPUs.

Beihang University (CN), Tsinghua University (CN)
Openalex Percentile: Top 27%
Advanced Data Storage Technologies
0.70
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

An eDRAM-Based In-Situ-Computing Processor for Homomorphic Encryption Evaluation on the Edge — Luchang Lei, Huazhong Yang, et al. · (2024) | TGRS Research Map | TGRS