rnd.farm

A stream of human generated randomness

updates

2016-02-11

Today the DNS for this experiment expired and the project is now offline. Thanks to all who participated.

2016-01-07

Read more about the project and its usage in this blog post: Evolutionary Failures (Part 1)

2015-02-17 - v2 released with data collection via websockets

A new version has been released, now collecting inherent randomness from each user's mouse/touch movements and key presses and their differences in timing frequency both in terms of interaction and network latency. This provides us with magnitudes more of raw collection data and less effort for you, the user. Additionally, all submitted numbers are both stored as-is and secondly are run through a chunking SHA-256 cryptographic hash function to distribute bits more uniformly. This is a similar approach as used by tools like TrueCrypt or Keepass. More details about this approach can be found at these links:

• Gathering entropy from mouse events
• Keepass
• TrueCrypt

WebSocket support

Whilst connected to the site (and if your browser supports WebSockets), you can see activity of all other current users. During recording your own random events, an histogram of collected bytes is displayed, giving an indication of the random distribution of your submitted source data.

Technical note: Since this histogram is byte based and submitted values have a variable bit length, the "00" bin (grouping bytes between 0x00 - 0x0f) will likely be the most dominant, since it will capture the truncated MSB end of the variable width integers. This has no impact on the SHA-256 digest these bytes are fed through. The original raw values are not split into bytes and stored as is (unsigned ints).

Non-WebSocket version

If your browser does not support WebSockets, you'll be automatically redirected to the previous, form-based version, here:

http://rnd.farm/form

why?

"Chance, predictability, and (true) randomness" is the theme of the 'ideas' section of Holo Magazine's upcoming second issue and we've been working on a Genetic Programming system to create a design for the cover and other aspects. The GP approach relies on selectively breeding programs by randomly generating, mutating and mating a population of hundreds of deeply nested abstract syntax trees over a potentially large number of generations. As you can guess, this all requires an endless amount of random numbers...

Even though we could (and so far did) use one of the many pseudo-random number generators readily available, Holo is also very much about documenting & involving a community which takes pride in experimenting with seeing things from alternative perspectives.

So in light of this:

We'd like to do a little experiment and see if a large (or even not so large) group of people can collectively generate better quality random numbers than an algorithm. Intuitively one would think and hope so, but I fear our sample set will be heavily biased. Yet, even with that the case, it will be fascinating to see if and how this might impact an evolutionary process as it consumes our random choices. At some point (once a sufficient amount of numbers has been collected/farmed - with your incredible help, of course! :), it should be quite enlightening to analyze the distribution itself in more detail.

how?

The initial site is merely targeted at kickstarting the collection of our random numbers (you can submit as many as you like).

This project was built with:

• Clojure / ClojureScript
• Leiningen
• Http-Kit
• Ring
• Compojure
• Hiccup
• Environ
• core.async
• clj-time
• thi.ng/domus
• thi.ng/color
• timbre

obtaining numbers

The site provides two simple means to get hold of submitted numbers via these routes:

/random[?n=1..1000]

Returns between 1 and 1000 of the submitted numbers in pseudo-random order in either JSON, EDN or CSV formats (specify via Accept header). For example:

curl http://rnd.farm/random?n=10 -H 'Accept: application/json'
[617716,151278150479,11,8585785,6513205632056302,8756875767,1500450271,505,3388227013,5]

curl http://rnd.farm/random?n=10 -H 'Accept: application/edn'
[3784939215633 32447100556 7845 1234568526536 10 4562144123 367689452 22 545626323265 48376218205]

curl http://rnd.farm/random?n=10 # CSV is default
4329847338943798,3,963,9,2445,56378,14415245255555457,1337,76,283901

/snapshot

This returns all submitted numbers in their original order as plain text (one number per line).

running locally

To start a local version on your own machine:

git clone https://github.com/postspectacular/rnd.farm.git
cd rnd.farm
RND_CONFIG=config.edn lein trampoline run

Alternatively, if you want to launch into a project REPL:

RND_CONFIG=config.edn lein repl

In the REPL you can reload the server like this:

(require 'rndfarm.handler :reload) (restart!)

After a few seconds, this will have launched the server on port 3000 and automatically opens the site in your browser. Numbers will be read & written to the files given in the config file (config.edn):

• SHA-256 digests are emitted in binary to the file given in config :digest :out-path.
• Raw ints are written as plain text to :raw :out-path

license

Distributed under the Apache Software License 2.0.

Copyright © 2015 Karsten Schmidt