Hacker News new | past | comments | ask | show | jobs | submit login

OK so what is random.org doing that they can claim it's a true random number generator (TRNG)? I looked at the page but didn't see any explanation, maybe they don't say, but as others have said there's no way it can be a TRNG. edit: I see the explanation on random.org

I'm not a mathematician or computer programmer but wouldn't a TRNG require infinite memory, infinite storage, infinite time to generate etc.? One number generated may be 2 but the next number may be negative infinity.

I work in a casino as a slot tech and sometimes even though I know it's not true some patrons can almost convince you there are patterns.




They use atmospheric noise to add entropy to their RNG. The definition of "True" in this case is rather loose, but once you get enough bits of entropy you can feed it into a PRNG or conditioners to give the output the qualities you want.

Here is some history of random.org http://www.random.org/history/

and the wiki page http://en.wikipedia.org/wiki/Random.org

the wiki page for atmospheric noise briefly discusses it's applications to "high quality" RNG http://en.wikipedia.org/wiki/Atmospheric_noise

Unfortunately, the source code to the RNG is close-source, and they say they aren't going to release it. But my guess is they use that random noise as a seed to a PRNG. Their FAQ has quite a lot of information though: http://www.random.org/faq/#Q1.2


> wouldn't a TRNG require infinite memory, ... etc.? One number generated may be 2 but the next number may be negative infinity.

I may be misunderstanding you, but true randomness doesn't mean 'a distribution that's non-zero over the whole integers' - those are orthogonal concepts. You can have a TRNG that gives you a random choice from just {0,1} - the size of the distribution isn't what makes it a TRNG, that just determines how much entropy you need.

(And a uniform distribution over the integers is impossibly no matter how good your RNG is, as it's mathematically undefinable).


>wouldn't a TRNG require infinite memory, infinite storage, infinite time to generate etc.? One number generated may be 2 but the next number may be negative infinity

What are you talking about here? Generating random Real numbers or something? First off you'd have to define a distribution of some sort or your goal is meaningless...but throw all that out. It provides random the sensible way, one bit at a time. Interpret that bit however you want. Memory requirements: zero (or maybe O(1) depending on design). Time requirement: O(n) where n is the amount of entropy you desire.


I guess what I'm asking is say I have a device that can generate a stream of numbers, integers, nothing fancy.

I want random integers with no limit to the size returned, a truly random number and by that I mean a true random integer without any limitation in size.

Really I guess I'm trying to grasp how a true random anything could exist or more to the point how a person could make a device or even be able to know if a random number or anything has been returned as a result.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: