Is spam a random event?
There's a lot of it, it's full of garbage, no one person controls it. It seems like it could make an interesting random number generator.
One of the old tricks for randomizing gameplay is to use the low bits off the analog inputs (or the disk seek times) to seed your random number generator. Seems like you could do the same sort of thing through statistical analysis of spam.
It's probably an intrinsically bad idea to allow an external source influence your security, but the idea of using spam to improve the security of commercial transactions seems deliciously ironic.
Thankfully, spam is not quite random. That's why current baysian spam filters work. Lovely idea, though. If it became useful, there would be services that will send spam by request. Hmmm... Can you get solicited unsolicited email?
Posted by: Yaron at October 21, 2004 11:23 AMWell, I was more wondering about the randomness of things like character count and MTBS (mean time between spam).
Posted by: Matt at October 21, 2004 11:54 AMMay be something there. Depends how much you really care about randomness. Things like Raymond Chen's graph show that size isn't entirely random, so you'll get result leaning towards the mean. As for MTBS (need to remember that one), I recall seeing a while ago statistics indicating a strong relation between the hour and amount of messages. Although both of these should be alright if you stick to the least significant parts... Could work I suppose, if you don't need a large amount of numbers in a small duration.
Posted by: Yaron at October 21, 2004 12:36 PM