Skip to content

ARK: Fully Homomorphic Encryption Accelerator with Runtime Data Generation and Inter-Operation Key Reuse

Authors: Jongmin Kim; Gwangho Lee; Sangpyo Kim; Gina Sohn; John Kim; Minsoo Rhu; Jung Ho Ahn
Year: 2022
Venue: MICRO 2022

type: accelerator collection: catalogue status: auto

Links: DOI · arXiv:2205.00922

Summary

Auto-extracted summary (unreviewed)

ARK is an ASIC accelerator for CKKS fully homomorphic encryption whose algorithm-architecture co-design removes the off-chip memory-bandwidth bottleneck of bootstrapping: evaluation keys and plaintexts are regenerated at runtime on chip and keys are reused across operations, shrinking the working set to fit on-chip memory. The microarchitecture adds an alternating data distribution policy and streamlined base-conversion, NTT and automorphism units to minimize on-chip data movement; the excerpt states that this drastically reduces the cost of HE operations including bootstrapping but gives no headline numbers.

Abstract

Homomorphic Encryption (HE) is one of the most promising post-quantum cryptographic schemes that enable privacy-preserving computation on servers. However, noise accumulates as we perform operations on HE-encrypted data, restricting the number of possible operations. Fully HE (FHE) removes this restriction by introducing the bootstrapping operation, which refreshes the data; however, FHE schemes are highly memory-bound. Bootstrapping, in particular, requires loading GBs of evaluation keys and plaintexts from off-chip memory, which makes FHE acceleration fundamentally bottlenecked by the off-chip memory bandwidth. In this paper, we propose ARK, an Accelerator for FHE with Runtime data generation and inter-operation Key reuse. ARK enables practical FHE workloads with a novel algorithm-architecture co-design to accelerate bootstrapping. We first eliminate the off-chip memory bandwidth bottleneck through runtime data generation and inter-operation key reuse. This approach enables ARK to fully exploit on-chip memory by substantially reducing the size of the working set. On top of such algorithmic enhancements, we build ARK microarchitecture that minimizes on-chip data movement through an efficient, alternating data distribution policy based on the data access patterns and a streamlined dataflow organization of the tailored functional units -- including base conversion, number-theoretic transform, and automorphism units. Overall, our co-design effectively handles the heavy computation and data movement overheads of FHE, drastically reducing the cost of HE operations, including bootstrapping.
Source: arxiv

Extracted details

Field Value Source Confidence
contribution runtime data generation to avoid loading evaluation keys and plaintexts from off-chip memory llm 0.70
contribution inter-operation key reuse to reduce the working set llm 0.70
contribution alternating data distribution policy based on data access patterns llm 0.70
contribution streamlined dataflow with tailored base conversion, NTT and automorphism functional units llm 0.70
contribution algorithm-architecture co-design targeting CKKS bootstrapping llm 0.70
schemes CKKS llm 0.70
libraries Lattigo llm 0.70
tasks FHE acceleration llm 0.70
tasks bootstrapping llm 0.70
hardware ASIC (proposed accelerator design) llm 0.70
security level 128-bit llm 0.70
interaction non-interactive FHE llm 0.70
code url https://github.com/tuneinsight/lattigo (URL found in the full text) regex 0.50
code url https://github.com/nucypher/nufhe (URL found in the full text) regex 0.50
llm relevance low llm 0.70

System card: ARK

Group 8. FHE hardware accelerators
Family ASIC (CKKS)
Arithmetic -
Interaction Hardware
Model scale -
Status Active
Relation to the SoK Generates data on the fly to reduce memory traffic.
Note Addresses the memory wall that the SoK identifies for LLMs.
Systems page ARK

Keyword mentions

Counts of vocabulary terms in the full text (first page of occurrence in brackets).

  • scheme: CKKS 25 [p.1], Leveled / somewhat HE 3 [p.1], FHEW 2 [p.16], MPC 2 [p.14]
  • library: Lattigo 10 [p.5], PALISADE 2 [p.18], HElib 1 [p.16], cuFHE / nuFHE 1 [p.18]
  • model: ResNet 24 [p.2], CNN 5 [p.2]
  • dataset: ImageNet 2 [p.12], MNIST 1 [p.11], CIFAR-10 1 [p.11]
  • hardware: GPU 23 [p.1], ASIC 13 [p.1], FPGA 12 [p.1], NVIDIA V100 2 [p.11], CUDA 2 [p.15], NVIDIA A100 1 [p.16], NVIDIA H100 1 [p.5], Intel Xeon 1 [p.12], AVX-512 1 [p.16]
  • technique: Bootstrapping 98 [p.1], NTT 40 [p.2], Key switching 18 [p.1], Rotations 14 [p.3], Rescaling 9 [p.3], RNS 9 [p.3], Convolution 6 [p.7], SIMD packing 2 [p.12]
  • assumption: LWE 4 [p.1], Lattice-based 4 [p.16], RLWE 2 [p.2]
  • topic: Homomorphic encryption 131 [p.1], Open source 3 [p.16]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/catalogue/8-115_ark-fully-homomorphic-encryption-accelerator-with-runtime-da.pdf catalogue 18 724.8 KB c1f2e01a2309

Database key: ark-fully-homomorphic-encryption-accelerator-with-runtime-data · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set ark-fully-homomorphic-encryption-accelerator-with-runtime-data COLUMN VALUE