Random Thoughts or How to Get True Random Numbers
4 Comments Published by tobias.escher June 5th, 2007 in *OIINEWS, all the restRight, so everyone with a first-year course in statistics knows that it is desirable to draw a random sample from a population in order to enable statistical inference. But how do you select really randomly?
Turns out that this is not as simple as it sounds. Most software “only” generates pseudo random numbers which means that ultimately the numbers are not really, well, random.
However, help is at hand so search no longer as random.org offers some simple and convenient ways of acquiring true random numbers, including for example a Perl module to access the service via a script. The random numbers are generated with the help of atmospheric noise which yields the desired properties of the random number distribution: uniform, non-deterministic and aperiodic.
There is a good overview about the problem on the site which also highlights additional sources of randomness (no, this blog is no official source as the topics of this blogs are not considered to be random in the strict mathematical sense…).
Ok ok, not that it would really matter if you are just drawing a few hundred cases from a population.
Search
About
Since October 2006 I am both a DPhil student as well as a research assistant at the Oxford Internet Institute and here I share with the accidental reader my musings on different aspects of the Internet and society. Feel free to comment or simply ignore :-)
-----------------------------------
Tobias Escher
Oxford Internet Institute
1 St. Giles
Oxford OX1 3JS
firstname.lastname@oii.ox.ac.uk
+44 (0)1865 287210
Thanks for a great post. One thing I don’t understand is how atmospheric noise would be considered random (in the sense that a machine/software-driven generator is not)? It seems like the same thing in regards to “true” randomness.
John Walker, retired Autodesk founder, offers hardware-generated random numbers as a free service– a gadget in his Swiss lab generates true random numbers by radioactive decay.
http://www.fourmilab.ch/hotbits/
Cheers, Ted
I don’t share the opinion that it’s really true random numbers, although as it’s stated in the webpage, any state is as possible as any other one. It isn’t completely impossible to reproduce the conditions (although very hard). Nevertheless, it’s quite hard to prove a number is really random.
There’s another nice site to generate random numbers (in this case, quantum based) in http://www.randomnumbers.info/
Best regards,
Sergio
Ted and Sergio,
thank you very much for your suggestions.
As I already said in my post, at the end of the day for the most things I would need random numbers for, it does not really matter whether they are true, almost true or just as good as true
.
What I do like however is the ability to interface with the random number generator over the Internet.
best, tobias