3 ms·
Coinflips sounds pseudorandom. Greying or mixing or whatever of pseudorandom or actual "trng" like static or radiation discharges is pretty well known. I.e. ta
by genewitch 5y ago
Coinflips sounds pseudorandom.
Greying or mixing or whatever of pseudorandom or actual "trng" like static or radiation discharges is pretty well known. I.e. take some source of randomness for 10000 bits, xor with each other, and use that to seed a cryptographically secure generator, topping up the seed as more bits become available, is pretty much the gold standard for "I need 50gbit bitstream of 'random'"
The other part seems to be the binning/bucketing/curve fitting, or "massaging" the output of that to only gives numbers in the range under scrutiny, which sounds like an implementation detail to me, rather than something that you can configure at ... Runtime? Compile time?
I'm sure there are real HRNG (or sensors for TRNG) that can approach ludicrous bitrates, I've never had the luxury of one, and I'm not entirely sure I could be convinced the output was any "better" than a vetted arrangement from my second paragraph.
All of this is to say: okay but we already do that and massaging randomness to give us results we want is what all this modeling stuff ... is, so other than some off in the weeds magnetic film manipulation, I'm assuming that's modelled, too. It's just models all the way down?
- hinkley 5y ago> xor with each other, and use that to seed a cryptographically secure generator Do not try to cook your moderate entropy data source. You can end up cooking what little entropy it had out of it. Many CSPRNGs allow seeds of arbitrary size. Just push all of your sketchy data into it, and let the hash function handle scavenging the entropy for you.
- johnisgood 5y agoJust because you said "psuedo" in all cases: it is "pseudo". Completely off-topic: do you live alone in the forest, or with others?
- genewitch 5y agoMy phone never fails to find new ways to embarrass. With others.
- johnisgood 5y agoIt is not embarrassing at all. :) I did not even want to comment on it initially, but then I thought I could help with it... That comment got down-voted already though. If you do not mind me asking: is it a community? I swear I have been wanting to live in a community such as Twin Oaks Community but I am in Europe! :(
- nick__m 5y agoEven tought the article is about true random number, pseudo random is more valuable for scientific simulations. In those simulations you don't typically want true randomness, you want a generator that pass a suite of randomness tests while having the ability to repeat that stream of numbers to replicate your experiment. Sometimes, you even discard numbers so different simulation runs with different parameters consume the same amount of numbers. A simplistic example: if rand()<.5 : doSomethingWithRand(rand()) else: doSomethingWithoutRand() rand()