Skip to content

EVA: An Encrypted Vector Arithmetic Language and Compiler for Efficient Homomorphic Computation

Authors: Roshan Dathathri; Blagovesta Kostova; Olli Saarikivi; Wei Dai; Kim Laine; Madanlal Musuvathi
Year: 2020
Venue: PLDI 2020

type: library collection: catalogue status: auto

Links: DOI · arXiv:1912.11951

Summary

Auto-extracted summary (unreviewed)

EVA is a general-purpose language and optimizing compiler for RNS-CKKS (targeting Microsoft SEAL) that hides scheme details by automatically inserting rescale, modswitch and relinearize instructions, deriving encryption parameters and rotation keys, and guaranteeing correctness and security at the default 128-bit level; a Python frontend lets programmers write image-processing applications in a few lines. EVA also serves as an intermediate representation for domain-specific compilers: re-targeting the CHET neural-network inference compiler onto EVA makes its generated programs 5.3x faster on average.

Abstract

Fully-Homomorphic Encryption (FHE) offers powerful capabilities by enabling secure offloading of both storage and computation, and recent innovations in schemes and implementations have made it all the more attractive. At the same time, FHE is notoriously hard to use with a very constrained programming model, a very unusual performance profile, and many cryptographic constraints. Existing compilers for FHE either target simpler but less efficient FHE schemes or only support specific domains where they can rely on expert-provided high-level runtimes to hide complications. This paper presents a new FHE language called Encrypted Vector Arithmetic (EVA), which includes an optimizing compiler that generates correct and secure FHE programs, while hiding all the complexities of the target FHE scheme. Bolstered by our optimizing compiler, programmers can develop efficient general-purpose FHE applications directly in EVA. For example, we have developed image processing applications using EVA, with a very few lines of code. EVA is designed to also work as an intermediate representation that can be a target for compiling higher-level domain-specific languages. To demonstrate this, we have re-targeted CHET, an existing domain-specific compiler for neural network inference, onto EVA. Due to the novel optimizations in EVA, its programs are on average 5.3x faster than those generated by CHET. We believe that EVA would enable a wider adoption of FHE by making it easier to develop FHE applications and domain-specific FHE compilers.
Source: arxiv

Extracted details

Field Value Source Confidence
contribution Encrypted Vector Arithmetic (EVA) language with Protocol Buffers serialization and Python frontend llm 0.85
contribution Optimizing compiler for RNS-CKKS inserting RESCALE/MODSWITCH/RELINEARIZE (waterline rescaling) and selecting parameters llm 0.85
contribution EVA as an IR for domain-specific FHE compilers llm 0.85
contribution CHET re-targeted to EVA with 5.3x average speedup llm 0.85
schemes CKKS llm 0.85
libraries SEAL llm 0.85
libraries CHET llm 0.85
libraries Protocol Buffers llm 0.85
tasks image processing llm 0.85
tasks neural network inference compilation llm 0.85
security level 128-bit llm 0.85
interaction non-interactive FHE llm 0.85
code url https://github.com/CEA-LIST/Cingulata (URL found in the full text) regex 0.50
code url https://github.com/homenc/HElib (URL found in the full text) regex 0.50
code url https://github.com/Microsoft/SEAL (URL found in the full text) regex 0.50
llm relevance low llm 0.85
key result speed-up: 5.3x on average (CHET neural network inference programs re-targeted onto EVA vs unmodified CHET) llm 0.85

System card: EVA / CHET

Group 7. FHE libraries, GPU back-ends and compilers
Family CKKS (SEAL)
Arithmetic Python/C++
Interaction Compiler/framework
Model scale -
Status Deprecated
Relation to the SoK Early CKKS compiler with automatic rescaling.
Note Archived but influential.
Code https://github.com/microsoft/EVA
Systems page EVA / CHET

Keyword mentions

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

  • scheme: CKKS 39 [p.1], BFV 11 [p.2], BGV 10 [p.2], MPC 8 [p.2]
  • library: EVA / CHET 218 [p.1], Microsoft SEAL 25 [p.2], nGraph-HE 10 [p.14], HElib 2 [p.14], ABY / ABY3 2 [p.15], EzPC / CrypTFlow 2 [p.15], PALISADE 1 [p.14], TensorFlow 1 [p.12]
  • model: LeNet 24 [p.12], Linear regression 1 [p.14]
  • dataset: MNIST 5 [p.12], CIFAR-10 5 [p.12]
  • hardware: Intel Xeon 1 [p.12]
  • technique: Rescaling 109 [p.1], Compiler 54 [p.1], Relinearization 30 [p.2], Rotations 17 [p.2], RNS 12 [p.1], SIMD packing 2 [p.1], Convolution 2 [p.16], Two-party computation 2 [p.15]
  • assumption: RLWE 2 [p.2]
  • topic: Homomorphic encryption 98 [p.1], Open source 7 [p.1]

Files

Path (under papers/) Collection Pages Size SHA-256 Note
references/catalogue/7-108_eva-an-encrypted-vector-arithmetic-language-and-compiler-for.pdf catalogue 16 581.2 KB 80f1171a20b6

Database key: eva-an-encrypted-vector-arithmetic-language-and-compiler-for-efficient · created 2026-11-05 · updated 2026-11-05 · edit with python3 tools/paperdb.py set eva-an-encrypted-vector-arithmetic-language-and-compiler-for-efficient COLUMN VALUE