Continuing on the final aspect of JamesT's very nice exploration of PiGI, we want to further examine the possibilities to create randomness from the output of a Geiger-counter. Of course we are using a PiGI as testbed for our experiments.

True randomness, in the sense of “provably unpredictable” is not easily available on a computer. Good arguments can be made that the available sources of entropy like clock jitter, floating analog inputs, network traffic etc, combined with the algorithmic magic of /dev/random and /dev/urandom are more than adequate for all non-cryptographic purposes, and probably unpredictable enough to safely use them in cryptography without worrying to much.

But paranoia and cryptography run hand in hand and using radioactive decay as source of entropy is an interesting application of physics to provide access to the elusive “true” randomness on a computer.

While the half-life gives the duration after which a particle has decayed with a probability of 50%, the exact moment of decay can not be predicted in any way. So a sample of radioactive material creates a stream of events in a Geiger-counter, which, due to the enormous amounts of atoms even in the tiniest sample, averages to a steady rate, but the specific event times (or delays between events) are unpredictable.

In the most fundamental form, our task is to produce a stream of equally probable 0's and 1's. To do this, we compare the last two delays between measured events. If one is longer than the other, the next bit is a 0 and if it is shorter it is a 1.

→ Read more...