There's this website called random.org, that says:

"RANDOM.ORG offers true random numbers to anyone on the Internet. The randomness comes from atmospheric noise, which for many purposes is better than the pseudo-random number algorithms typically used in computer programs."

I was wondering... If I could get data every time my cat moved his head, that would seem pretty much random to me . Cats are random, and, therefore, might be a nice input. (Sorry for the random topic, by the way...)

http://thecatapi.com/

http://placekitten.com/

Sure, why not?

Silicon Graphics once used pictures of a lava lamp to seed their RNG in some project: http://en.wikipedia.org/wiki/Lavarand

Lava? haha.. Interesting how the choices for achieving randomness are as random as the results. Cats might be much more "accessible". One could figure out how to collect the data from the movement of the cat's head and store it in some kind of stack (Supposing that, at some point, the cat stops moving its head while sleeping or whatever (improbable)), and one could easily use more than one cat as an input (Actually, the same principle could be applied to other animals, or anything more efficient like insects or whatever).

Why don't CPUs all have a RNDSEED instruction built in?

Intel actually did have a hardware RNG in their chipsets: http://www.cryptography.com/public/pdf/IntelRNG.pdf

Though I guess no-one used it, seems they have removed it now.

Love the Cat RNG btw, could be a fun weekend hack, use one of those cheap micro controllers, preferably with some kind of wireless transfer, then add an accelerometer and a headstrap for the cat.

Then you could collect it over the internet for a massivly distributed cat rng :)

Intel actually did have a hardware RNG in their chipsets: http://www.cryptography.com/public/pdf/IntelRNG.pdf

Though I guess no-one used it, seems they have removed it now.

Thanks for the link, I didn't know about this RDRAND instruction. According to a quick google, it's still supported and is an official part of the intel x86/x64 specs now.

When I worked in gambling-games, we were required by law to use true-RNGs (though we could do QA & brute-force/montecarlo proofs using PRNGs). I didn't know much about our hardware, but I think it was based on radioactive decay rather than cats ;-)

There was some controversy on RDRAND because there was suspicion that it may have been backdoored. As you all know, NSA conspiracies are in vogue these days, and IC's only a few dozen nanometres across built using private manufacturing processes are not the easiest thing to study. It is probably best to only use it as just another source to the operating system's entropy pool*, however, instead of using it directly, as I believe the Linux and BSD kernels already do, to make sure it cannot poison software on its own. I kinda feel like it should have been made a privileged instruction, to ensure it can only be consumed by the operating system (which is really the only service you can trust to provide cryptographically secure random numbers under normal circumstances and without special hardware, and even then, unless your software has very specific requirements e.g. certified online gambling server, letting that hardware interface with the operating system's entropy pool* is the recommended procedure).

* unless you are running Windows, in which case it is recommended to trust only yourself (tongue in cheek.. or.. is it...)

According to a quick google, it's still supported and is an official part of the intel x86/x64 specs now.

Oh, great! :)

My sources for it being removed was wikipedia, I guess it was just removed from the chipset and moved into the cpu.

A analog noise generator never has flat/equal response but has true randomness.