QuietOT Implementation 🤫

QuietOT (to appear at AsiaCrypt 2024) is a lightweight oblivious transfer extension protocol with a non-interactive public-key setup. This implementation is intended to benchmark the performance of QuietOT, and can be used to reproduce a subset of the results reported in Table 2 of the evaluation section.

Organization

Directory
quiet-bipsw/	Implementation of QuietOT with the BIPSW wPRF.
quiet-gar/	Implementation of QuietOT using the GAR wPRF.
bcmpr-bipsw/	Implementation of the BCMPR PCF using the BIPSW wPRF.
other/	Partial implementations of the BCMPR and OSY pseudorandom correlation functions for benchmarking purposes.

Dependencies

• OpenSSL
• GNU Make
• Cmake
• Clang

Installing dependecies on Ubuntu and CentOS

Install dependencies (Ubuntu):	Install dependencies (CentOS):
sudo apt-get install build-essential	sudo yum groupinstall 'Development Tools'
sudo apt-get install cmake	sudo yum install cmake
sudo apt install libssl-dev	sudo yum install openssl-devel
sudo apt install clang	sudo yum install clang

Installing dependencies on MacOS

On MacOS, use homebrew to install dependencies. cmake and clang can be installed via xcode-select --install. OpenSSL can be installed via brew install openssl or manually.

brew install openssl
xcode-select --install  # Installs cmake and clang

Note: If installing OpenSSL manually on Mac, then you will need to update the library path in the Makefiles to point to the correct directory. Change this -I/opt/homebrew/opt/openssl/include to -I[PATH]/openssl/include

Troubleshooting (MacOS)

If you see "library 'crypto' not found", try:

1. Check if OpenSSL is properly installed: brew list openssl
2. Verify the OpenSSL path: brew --prefix openssl

Hardware Requirements

Memory: At least 4GB RAM recommended.

Processor:

• Basic version works on any x86_64 processor.
• AVX512 acceleration requires Intel processor with AVX512 support.

Expected performance:

• With AVX512: ~1,200,000 OT/s.
• Without AVX512: ~500,000 OT/s on Linux.
• Without AVX512: ~1,200,000 OT/s on M1 Mac.

Quick Start

To get QuietOT up and running quickly:

git clone https://github.com/sachaservan/QuietOT
cd QuietOT
cd quiet-bipsw
make && ./bin/test

Running tests and benchmarks

Test and benchmark for QuietOT:

cd [quiet-bipsw | quiet-gar]
make && ./bin/test 

By default, the benchmarks are compiled to generate $2^{20}$ OTs. You can change the number of OTs generated (for QuietOT benchmarks only; not implemented for the BCMPR benchmarks) by running

make NUM_OTS=[log2 number of OTs]

For example, make NUM_OTS=18 will generate $2^{18}$ OTs. Make sure to run make clean before recompiling.

Note: trying to generate too many OTs (e.g., setting NUM_OTS=25 will cause performance issues and may cause the benchmarks to stall or crash depending on how much memory you have on your hardware).

Running BIPSW with AVX512

Note: requires AVX512 hardware support! (e.g., does not work on M1 Mac)

cd quiet-bipsw
make AVX=1 
make && ./bin/test 

Benchmarking BCMPR (with BIPSW wPRF):

cd bcmpr-bipsw
make && ./bin/test

Benchmarking BCMPR (with GAR wPRF) and OSY:

cd other
make && ./bin/bench

Reproducing Table 2 of the paper

Simply run the benchmarks as described above on the respective hardware. The output of each benchmark will look something like this:

Took 871.88 ms to generate 1048576 OTs
PASS
...
Took 931.52 ms to generate 1048576 OTs
PASS

******************************************
SUMMARY
Avg. time: 8892.60 ms to generate 1048576 OTs
Performance: 117915.54 OTs/sec
Number of trials: 10
******************************************

Notes:

• Default NUM_OTS=20 produces the OTs/sec values in the table on the respective hardware.
• Public key sizes in the table are derived in the full version of the paper.
• Bits/OT are derived analytically in the paper by analizing the ring size.

Future Improvements

• Implement the public-key setup for QuietOT
• Find a way to use SIMD instructions for the GAR variant
• Implement SIMD for ARM architectures

Citation

@inproceedings{QuietOT,
  author       = {Geoffroy Couteau and
                  Lalita Devadas and
                  Srinivas Devadas and
                  Alexander Koch and
                  Sacha Servan-Schreiber},
  title        = {{QuietOT}: Lightweight ObliviousTransfer with a Public-Key Setup},
  note         = {\url{https://eprint.iacr.org/2024/1079}},
  url          = {https://eprint.iacr.org/2024/1079}
  editor       = {Kai-Min Chung and
                  Yu Sasaki},
  booktitle    = {Advances in Cryptology - {ASIACRYPT} 2024 - 30th
                  International Conference on the Theory and
                  Application of Cryptology and Information Security,
                  Kolkata, India, December 9-13, 2024
                  },
  publisher    = {Springer},
  year         = {2024},
}

Acknowledgements

We use the super fast Polymur hash for universal hashing. We thank Maxime Bombar for reviewing and providing feedback on the code.

⚠️ Important Warning

This implementation is intended for research purposes only. The code has NOT been vetted by security experts. As such, no portion of the code should be used in any real-world or production setting!