drb-client 0.3.0

Distributed Randomness Beacon client

drb-client

Distributed Randomness Beacon client

Gathers entropy from multiple drand instances, securely mixes responses and outputs to kernel entropy pool or stdout. Suitable for use with League of Entropy servers (see "Running" section).

drb-client can be used as a standalone source of high-quality random number, or as additional source for entropy pool in Linux kernel. Produced amount of entropy should be sufficient to derail attacks based on RNG predictability.

Cryptography background

drb-client polls list of drand servers with given interval and requests private entropy data portion from each one. Communication between drb-client and drand is protected with AES256-GCM encryption derived on BN-256 curve on top of TLS (this is imposed by drand API format).

drb-client constructs each output of entropy from at least quorum (-Q option) distinct inputs. It is assumed at least node_count - quorum + 1 nodes produce truly unpredictable secure random numbers, so any quorum of distinct responses definitely contain at least one truly random input (due to pigeonhole principle).

Entropy portions from beacon servers are mixed using stateful HKDF-based mixer. Each sufficient set of random responses is used to produce random output and new salt value for HKDF mixer. Therefore, mixer output becomes unpredictable after successful generation of first output, even if all beacon servers get compromised and start feeding client with biased data.

Default poll interval is 60 seconds and such interval is chosen for a reason. drand generates entropy for each response using its /dev/urandom. On Linux urandom gets reinitialized from /dev/random each 1 minute. So there is no reason to fetch random data more often: responses between reinitializations are in functional dependence.

Installation

Requires Python 3.5.3+

From PyPI

pip3 install drb-client

From source

Download or clone repository and run following command in source code directory:

pip3 install .

Running

There are few available entropy sinks (option -O):

• devrandom - (default) writes collected entropy into /dev/random device, without increment of kernel counter of available entropy in pool.
• stdout - writes collected entropy into standard output.
• rndaddentropy - writes collected entropy into /dev/random device with increment of kernel counter of available entropy in pool. Requires superuser privileges to operate.

You may obtain latest group.toml config with list of League of Entropy servers here.

Run with default sink:

drb-client group.toml

Contribute into /dev/random and increment entropy counter:

sudo drb-client -O rndaddentropy group.toml

For logging into file see "Synopsis" section.

Synopsis

$ drb-client --help
usage: drb-client [-h] [-v {debug,info,warn,error,fatal}] [-l FILE]
                  [-Q QUORUM] [-T PERIOD] [-w TIMEOUT]
                  [-O {stdout,rndaddentropy,devrandom}]
                  group_config

Distributed Randomness Beacon client

positional arguments:
  group_config          group config

optional arguments:
  -h, --help            show this help message and exit
  -v {debug,info,warn,error,fatal}, --verbosity {debug,info,warn,error,fatal}
                        logging verbosity (default: info)
  -l FILE, --logfile FILE
                        log file location (default: None)

poll options:
  -Q QUORUM, --quorum QUORUM
                        minimal answers required on each poll. Default value
                        is (node_count // 2 + 1). (default: None)
  -T PERIOD, --period PERIOD
                        poll interval for each source (default: 60)
  -w TIMEOUT, --timeout TIMEOUT
                        timeout for each request (default: 4)

output options:
  -O {stdout,rndaddentropy,devrandom}, --output {stdout,rndaddentropy,devrandom}
                        entropy output (default: devrandom)

Deployment

You may obtain latest group.toml config with list of League of Entropy servers here.

See deploy/drb-client.service for example of systemd unit file.

Credits

• League of Entropy project
• drand project
• Jack Lloyd for BN256 implementation for Python