MD-ML

This is the repo of the paper MD-ML: Super Fast Privacy-Preserving Machine Learning for Malicious Security with a Dishonest Majority. We are currently engaged in extensive efforts to refactor our code and to write the documentation.

Building

Dependencies

The code has been tested on Windows 11, Ubuntu 22.04, and macOS 14.5. Building the code requires the following dependencies:

A C++20-compatible compiler
- GCC-10 or later, Clang-10 or later, latest version of Apple-Clang.
- MSVC compiler is not supported, we recommend MinGW-w64 on Windows.
CMake (3.12 or later)
The Boost Library (1.70.0 or later)
The Eigen Library (3.0 or later)

On Ubuntu, you can install all the dependencies via:

sudo apt install build-essential cmake libboost-system-dev libeigen3-dev

On macOS, you can install them using HomeBrew:

xcode-select --install # install the command-line tools
brew install cmake boost eigen

On Windows, the MSVC compiler is not supported, please use gcc or clang. We recommend using MinGW-w64. The steps for configuration is more complicated. Install MinGW-w64 and CMake. Then download the source code of Boost and Eigen. To install Eigen, follow the installation instructions "Method 2. Installing using CMake", you may need -G "MinGW Makefiles" option when invoking cmake. For Boost, just unpack the code, no building is required. Finally, Add the path of Boost and Eigen to environment variable PATH.

Building

First clone the project and create the build directory:

git clone https://github.com/NemoYuan2008/MD-ML.git
cd MD-ML
mkdir build && cd build

On Linux and macOS, configure and build the project with:

cmake -DCMAKE_BUILD_TYPE=Release ..
cmake --build .

On Windows with MinGW-w64, you might need to specify -G "MinGW Makefiles" option when invoking cmake:

cmake -G "MinGW Makefiles" -DCMAKE_BUILD_TYPE=Release ..
cmake --build .