Single & Multi Key Homomorphic Encryption Library

This C++ library implements the Full-RNS CKKS Homomorphic Encryption scheme. Apart from the basic use-case of CKKS, the library also implements the multi-key variant explained in this paper. The library doesn't implement Galois keys and can be used as a reference for future C++ multi-key CKKS libraries. The library was created as part of final-year university project in order to compare with existing HE libraries while using HE for contact tracing.

⚠️ It is an experimental library and should not be used in production.

Build

The library depends on Protobuf and BigInt. While building BigInt is included in CMake, Protobuf needs to be installed beforehand.

The command for building is:

1. cmake -B build .
2. cmake --build build

After this, the library will be created inside the build folder.

Build for iOS/Android

The library can also be compiled for iOS/Android.

iOS

Protobuf needs to be built firstly. Using their source code, we can generate a binary with CMake and Xcode. After this, the smkhe library can be built.

1. Generate the Xcode project with:

   cmake -S . -B build-ios -GXcode -DCMAKE_SYSTEM_NAME=iOS "-DCMAKE_OSX_ARCHITECTURES=arm64;x86_64" -Dprotobuf_BUILD_TESTS=OFF

2. Then, using the project created inside build-ios, the protobuf binary can be built.

3. Modify the PROTOBUF_LIB variable from line 6 of the CMakeLists file to point to the binary

4. Use CMake with the CMakeLists file from the cmake folder as in the first step to generate the iOS project of smkhe

Android

1. Using the details from here, we can build the Protobuf binary using CMake
2. Modify the PROTOBUF_LIB variable from line 6 of the CMakeLists file to point to the binary
3. Similarly to 1, build the smkhe binary using CMake

References

1. Homomorphic encryption for arithmetic of approximate numbers
2. Efficient multi-key homomorphic encryption with packed ciphertexts with application to oblivious neural network inference
3. A full rns variant of approximate homomorphic encryption
4. Faster homomorphic discrete fourier transforms and improved fhe bootstrapping