msh086 / FDFB-TCHES-2024

The code for Fast and Accurate: Efficient Full-Domain Functional Bootstrap and Digit Decomposition for Homomorphic Computation, https://doi.org/10.46586/tches.v2024.i1.592-616

Invoking FDFB and Homomorphic Decomposition

Building

First build and install OpenFHE using the following commands. Replace YOUR_CMAKE_INSTALL_PREFIX with your own path.

git clone https://github.com/openfheorg/openfhe-development.git
cd openfhe-development
git checkout 745a492
git apply 20230504.patch
cmake -S . -B build -DCMAKE_BUILD_TYPE=Release -DWITH_NATIVEOPT=ON -DCMAKE_INSTALL_PREFIX=YOUR_CMAKE_INSTALL_PREFIX
cmake --build build -j8 && cmake --install build

Google Benchmark is also required if you want to meaure the performance of FDFB and homomorphic decomposition algorithms.

Modify the CMakeLists.txt in the current directory (not OpenFHE directory) by replacing YOUR_CMAKE_INSTALL_PATH with the path your just passed to CMake. If Google Benchmark is unavailable, comment the corresponding lines in CMakeLists.txt and remove benchfdfb and benchsign from target executables.

Now run the following commands.

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j8

The executables can be found in the build directory.

Running

benchfdfb benchmarks all FDFB algorithms. benchsign benchmarks all homomorphic decomposition algorithms. Their CLI is managed by Google Benchmark.

evalfunc evaluates FDFB. evalsign performs homomorphic decomposition. evalrelu evaluates large-precision ReLU with $\textbf{HomDecomp-Reduce}$. The parameter set used by the three executables can be specified using command line arguements. A basic usage is printed when invoking these executables without any argument. Refer to their source code for a full list of predefined parameters sets.

Parameter Selection and Noise Analysis

param.py provides convenient functions to estimate the noise of FDFB and homomorphic decomposition algorithms. To use these functions, type from param import * in Python command line. Refer to the python file for details on the usage of these functions.

sage_code.txt provides a simple template to measure the concrete security of an LWE (or RLWE) scheme. Both SageMath and lattice-estimator are required to use it.