This repository contains the code used to generate Tables 1--4 in [MP24].
Experimental data in Tables 1 and 2 was obtained using the provided HElib code [HElib] (with HElib version 2.2.1). Experimental data in Tables 3 and 4 was obtained using the provided Microsoft SEAL code [SEAL] (with SEAL version 4.0). Our noise growth estimates in Tables 1--4 were obtained using a python script.
Heuristics
The python script generate_bgv_heuristics_tables.py generates the noise growth estimates reported in the columns [CLP20] and Ours in Tables 1--4. It is best run using SageMath [SAGE].
Within Sage:
load("generate_bgv_heuristics_tables.py")
HElib
The HElib files BGV_clp20.cpp (for Table 1) and BGV_deep.cpp (for Table 2) were developed to run with HElib (version 2.2.1). With that version of HElib installed, add the folders BGV_CLP20 and BGV_deep to the folder HElib/examples/. These files can then be compiled and run as for the other HElib examples.
In /HElib/examples:
cmake .
make
Then in /HElib/examples/bin:
./BGV_deep
Note that the files BGV_clp20.cpp and BGV_deep.cpp require a slight modification to the Ctxt class, namely that the Ctxt::tensorProduct() function is made public.
SEAL
The provided files 4_bgv_basics_CLP20.cpp (for Table 3) and 4_bgv_basics_bgv_deep.cpp (for Table 4) were developed to run with SEAL (version 4.0). With that version of SEAL installed, they can be swapped in for the file 4_bgv_basics.cpp in the SEAL examples (SEAL/native/examples) and compiled and run as for the original SEAL examples.
In SEAL/
cmake -S . -B build -DSEAL_BUILD_EXAMPLES=ON
cmake --build build
In /SEAL/build/bin
./sealexamples
[CLP20] Anamaria Costache, Kim Laine, Rachel Player. Evaluating the effective- ness of heuristic worst-case noise analysis in FHE. In ESORICS 2020. Preprint available at: https://eprint.iacr.org/2019/493
[HElib] HElib. https://github.com/homenc/HElib
[MP24] Sean Murphy, Rachel Player. IACR Communications in Cryptology. Volume 1, No. 2, 2024. Preprint available at: https://eprint.iacr.org/2019/452.pdf
[SAGE] https://www.sagemath.org
[SEAL] Microsoft SEAL (release 4.0). Microsoft Research, Redmond, WA. March, 2022. https://github.com/microsoft/SEAL