HELIOPOLIS

HELIOPOLIS: Verifiable Computation over Homomorphically Encrypted Data from Interactive Oracle Proofs is Practical

Paper

eprint 2023/1949

@misc{cryptoeprint:2023/1949,
author = {Diego F. Aranha and Anamaria Costache and Antonio Guimarães and Eduardo Soria-Vazquez},
title = {HELIOPOLIS: Verifiable Computation over Homomorphically Encrypted Data from Interactive Oracle Proofs is Practical},
howpublished = {Cryptology ePrint Archive, Paper 2023/1949},
year = {2023},
note = {\url{https://eprint.iacr.org/2023/1949}},
url = {https://eprint.iacr.org/2023/1949}
}

Requirements

• Basic development packages (On Debian-based systems: sudo apt install make gcc cmake)
• Python 3 (Tested on Python 3.10.12)
• GCC (Tested on GCC 11.4.0)
• Strongly recommended: A processor with AVX-512 IFMA (Tested on Intel i7-1165G7 and Intel Xeon 8375C). We have added a compilation option that does not require it, but please be aware that it is mostly unoptimized and may be more unstable.

Running

1) Clone this repository

git clone https://github.com/antoniocgj/HELIOPOLIS.git

2) Run python3 test_fri.py -h to see the following help text:

usage: test_fri.py [-h] [-i INPUT_SIZE] [-e EXPANSION_FACTOR] [-m NUM_COLINEARITY_TESTS] [-t THREADS]

Evaluates FRI over encrypted polynomials

options:
  -h, --help            show this help message and exit
  -i INPUT_SIZE, --input_size INPUT_SIZE
                  Log_2 of the degree bound of input polynomials (default=7)
  -e EXPANSION_FACTOR, --expansion EXPANSION_FACTOR
                  Expansion factor (1 / \rho) (default=2)
  -m NUM_COLINEARITY_TESTS, --colin_tests NUM_COLINEARITY_TESTS
                  Number of colinearity tests per round of FRI (default=102)
  -t THREADS, --threads THREADS
                  Number of threads for the prover (default=2)

The Python code will compile the C library during the first execution.

To recompile the code (in case you want to run it in another machine, for example), delete the build:

rm -rf c_lib/lib/libfric.so
rm -rf c_lib/src/third-party/hexl/build/

Execution examples:

1. Parameter set FRI₀ with 4 threads for input of size 2¹¹:

python3 test_fri.py -i 11 -e 2 -m 102 -t 4

2. Parameter set FRI₃ with 4 threads for input of size 2⁷:

python3 test_fri.py -i 7 -e 16 -m 26 -t 4

[!WARNING]
The optimized version of this implementation requires AVX-512 IFMA. It should be significantly slower and possibly unstable without it (but it should still work).

Code structure

• test_fry.py: main script, it instantiates and runs FRI.
• friC.py: ctypes wrapper to call optimized (C/CPP-implemented) functions
• other Python scripts: high-level implementations of HE-FRI and all subprocedures required to implement it (Merkle commitments, NTTs, Fiat-Shamir...).
• c_lib/src: optimized C/CPP subroutines used by the Python scripts. It also contains a fully optimized version of HE-FRI's verification (c_lib/src/verifier.cpp).

Implementation changelog

Compared to the first version of the paper:

• Removed dependency on HELib (Replaced with HEXL)
• Removed dependency on Sagemath
• Readjusted HE parameters to make the prover faster at a small cost for the verifier.
• Added a multi-threaded version for the folding and batching operations.

License

Apache License Version 2.0

Adapted from the FRI implementation of Stark Anatomy - Apache License Version 2.0 - Copyright Alan Szepieniec

Additionally, this repository contains code from:

• MOSFHET: Apache License Version 2.0 - Copyright Antonio Guimarães et al. - See their detailed copyright information.
• Intel HEXL: Apache License 2.0 - Copyright 2020 Intel Corporation
• BLAKE3: Apache License 2.0 - Copyright 2019 Jack O'Connor and Samuel Neves