Maybe it’s the caffeine…

I just had an interesting idea:

You all know that random number generators are not truely random, as they generally get their input from something that could theoretically be predicted (temperature, radio background noise, etc).

I think it would be interesting to see if a person could be used as a random number generator. The data could be fairly easily gathered by sitting a person down at a speech recognition device that will respond to ‘one’ and ‘zero’ or ‘yes’ and ‘no’, recording them as bits in a key. If the person were to rattle off 1024 yes/no or one/zero answers, then you would have a 1k key. My presumption is that if you told a person to randomly pick one of the possible answers, they would tend to get fairly random after a while. But, it would be interesting to see. Standard tools for analyzing randomness of data could be used to see just how truely random the person is. It wouldn’t be hard to collect a number of high-bit samples from an individual, nor would it be difficult to collect samples from any number of people, allowing for proper analysis of a human-based random number generator. It’d also be an especially geeky way to generate PGP/GPG keys.