Distributed Verifiable Random Functions: an Enabler of Decentralized Random Beacons

This code repository provides a C++ implementation of three Distributed Verifiable Random Functions (DVRF) described in Fetch.AI's technical report Fully Distributed Verifiable Random Functions and their Application to Decentralised Random Beacons and named respectively GLOW-DVRF, DDH-DVRF and Dfinity-DVRF. A Distributed Verifiable Random Function is a particular case of secure multiparty computation that allows a set of mutually distrustful servers to initialize and compute a Verifiable Random Function f_sk(x) for strings x, with no trusted central party. A Verifiable Random Function can be seen as the public-key cryptography version of a keyed cryptographic hash. A prominent application for such protocols is in PoS-based consensus in distributed ledgers which require a reliable, unpredictable and unbiased source of entropy or Decentralized Random Beacon (DRB) to select block producers.

Disclaimer ❗

Fetch.AI makes no representation or guarantee that this software (including any third-party libraries) will perform as intended or will be free of errors, bugs or faulty code. The software may fail which could completely or partially limit functionality or compromise computer systems. If you use or implement this library, you do so at your own risk. In no event will Fetch.AI be liable to any party for any damages whatsoever, even if it had been advised of the possibility of damage.

As such this codebase should be treated as experimental and does not contain all currently developed features.

This code is licensed under the Apache software license (see LICENSE file). Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

Further implementation details are available for interested readers.

Summary ⭐

This code base implements two new DVRF (and DRB) designs named GLOW-DVRF (and GLOW-DRB) and DDH-DVRF (DDH-DRB), whose high-level specification can be found in the technical report [1]. GLOW-DVRF uses Pairing-Based Cryptography, while DDH-DVRF uses standand Elliptic Curve Cryptography. This code also implements the pairing-based Dfinity-DVRF (and Dfinity-DRB) outlined in [2].

The report defines two security properties for DVRFs, named standard and strong pseudorandomness. It is shown that both new designs GLOW-DVRF and DDH-DVRF meet strong pseudorandomness, while Dfinity-DVRF meets standard pseudorandomness. Furthermore, our experiments show that both GLOW-DVRF and DDH-VRF outperform Dfinity-DVRF in running time (for details take a look at the figures shown next and the report).

DVRF protocols and associated properties

Running time per node for each protocol to generate one round of entropy for a DRB with a total of 50 nodes

Running time per node for each protocol to generate one round of entropy for a DRB with a total of 200 nodes

Running times have been obatined on a laptop running Ubuntu 18.04. LTS with 64bit Intel Core i7-8550U processor, with 4GHz capacity, and 16GiB of memory.

Getting Started ⚒️

This project is built using CMake and requires the following dependences

GNU Multiple Precision Arithmetic Library (GMP)
Protocol Buffer
OpenSSL
Libsodium
POSIX Threads
pkg-config

On Ubuntu 🐧

On Ubuntu 19.10

Update and install dependencies

sudo apt-get update
sudo apt-get install cmake libgmp-dev libssl-dev libpthread-stubs0-dev pkg-config libprotobuf-dev protobuf-compiler libsodium-dev

On Ubuntu 18.04

Update and install dependencies

sudo apt-get update
sudo apt-get install cmake libgmp-dev libssl-dev libpthread-stubs0-dev pkg-config

To obtain the latest versions of the other dependencies please follow the instructions for Libsodium and Protocol Buffer.

On Mac 🍎

Install brew and then dependencies

/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
brew install cmake gmp protobuf openssl libsodium pkg-config

Building the Project

Once all dependences have been installed clone the repository and initialise submodules

git clone https://github.com/fetchai/research-dvrf.git
cd research-dvrf
git submodule update --init --recursive

Build the project

mkdir build
cd build
cmake ..
make

For Mac users, problems with finding OpenSSL might be remedied by specifying the root directory path to cmake. This can be done by replacing cmake .. with

cmake -DOPENSSL_ROOT_DIR=/usr/local/Cellar/openssl/[version]/ ..

Run tests (from build directory)

./lib/test/consensusTest

Note, the network nodes on the benchmarks described below run on local host with port numbers starting at 1024.

Benchmarks 🏭

PartialEval and Combine Functions

The two functions, PartialEval and Combine, used in generating each randomness value can be benchmarked for all implementations using Catch2. To do this, run the following command from the root directory

bash scripts/benchmark_functions.sh

which will prompt for the name of the file to write the benchmarks to. This runs the pairing-based protocols with the curves in order BN256, BLS12-381, and BN384.

To select a particular implementation out of [dfinity], [glow] and [ddh] run, for example,

./lib/benchmarks/consensusBenchmarks "[dfinity]"

Similarly, a specific benchmark can be run by replacing [dfinity] with the name of the benchmark. The function Combine is dependent on N, and is bencharked for N = 50, 100, and 200 with threshold N/2 + 1. Average times for computing a single random value is then obtained by combining the benchmarks for PartialEval and Combine.

Random Beacon

This app benchmarks the decentralised random beacon for local or network nodes for a chosen number of rounds. The benchmarking app for the decentralised random beacon circumvents the DKG step by distributed pre-comuted keys to all nodes allowing faster benchmarking of the DRB for a large number of nodes. The app outputs the following benchmarks:

Total time: time for all nodes to complete a fixed number of random beacon rounds
Average unite time/round: time for a node to complete one round of the random beacon

Example with N = 15 local nodes with zero latency, threshold T = 8 and r = 10 rounds

./apps/benchmark_threshold_signing/BenchmarkThresholdSigning -N 15 -T 8 -r 10

To view the app options and default parameters

./apps/benchmark_threshold_signing/BenchmarkThresholdSigning --help

Randomness Generation 🎰

To generate a series of random values using the available implementations run, from the build directory, the random beacon app for the required number of rounds

./apps/random_beacon/RandomBeacon -r 10

The random value is obtained by computing the SHA512 hash of the group signature and is displayed in hexadecimal format. This app bypasses the DKG using a trusted dealer, and can therefore start the generation of random values for a large number of participants quickly. By default, the mcl implementation of the GLOW-DVRF, with curve BN256 and 10 local nodes, is run. To view the other default values and command line interface to change them use --help.

Further implementation details are available here.

Contributing

We welcome any contributions to this repository. If you would like to contribute, please make changes on a fork of the repository and submit a pull request for review and merging.

References 🎓

[1] Galindo, D., Liu, J., Ordean, M., and Wong, J. (2020) Fully Distributed Verifiable Random Functions and their Application to Decentralised Random Beacons.

[2] Hanke, T., Movahedi M., and Williams, D. (2018) DFINITY Technology Overview Series, Consensus System.

farazshaikh / research-dvrf