Timeline for How fast does a pseudorandom number generator have to be in order to be competitive?
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Sep 29, 2018 at 17:24 | comment | added | James Huddle |
1dfbbfb05c61b06846f33be25ecfbba7 001123345566678aBBBBBBBccdeeffff (7,4) 68e9a9046e6c510367e88c733178a34f 001133334456666777888899aacceeef (4x3) 62e7174343ce637b13a80d2a0786557b 001122333344556667777788aabbcdee (5,4) 540447db380be3bb65e8ab0b4771d353 000133334444555677788aBBBBBBddee (6,4x2) 62ab103319834a906e82ecaf21bd10e2 000111122223334668899aaabbcdeeef (4x2) 28c9b1db46f466766368e329608ffebe 012233446666666788899bbbcdeeefff (7)
|
|
| Sep 29, 2018 at 17:24 | comment | added | James Huddle | In my defense, I cranked out a simple 128-bit PRNG using glibc's rand() and concatenating 4 of them and printing. "noticeably" fewer repeats. Why? not sure. Mine just seems more random, and hence, harder to crack. If you want "math", you could always "do the math" and extrapolate from my "31-out-of-32" example, which is real math. | |
| Sep 28, 2018 at 23:03 | comment | added | James Huddle | I agree that seeing "34" like that in two outputs would be "reading a lot into it." But if you see "34" in 6 out of 6 outputs, I'd say you've reduced your brute force effort by a factor of 100. Of course, you're betting that the 7th will have a "34" and it's always possible that it was just pure randomness (I understand the counterintuitive qualities of pure random values). I'm just saying that if I were designing a PRNG system, I would work hard to avoid such coincidences. And yes, I know how naive that sounds. Sorry about the lack of math proof. | |
| Sep 28, 2018 at 21:58 | comment | added | Sophie Swett | If you're suggesting that OpenSSL's PRNG is inadequate because it shows patterns, then this answer needs to be improved by including an analysis showing that the patterns are statistically significant. | |
| Sep 28, 2018 at 20:33 | comment | added | iheanyi | You're falling into a trap. I recall a complaint that consecutive second factor codes repeated a few digits and so it was not truly random. Something like "182345" and "405346" - there's a repeat of "34". Just because something stands out to you does not mean it is not random. Humans are great at seeing patterns including the ones that don't exist. | |
| Sep 28, 2018 at 20:05 | comment | added | poncho | I believe that a truly random output would show the same artifacts that you complain about OpenSSL; for example, a truly random 16 byte output would have 9 or repeated hex digits once every circa 370000 outputs. So, yes, if you go through enough of them, you would expect to see such extreme cases (and 'enough' needn't be that many...) | |
| Sep 28, 2018 at 18:30 | review | First posts | |||
| Sep 28, 2018 at 22:56 | |||||
| Sep 28, 2018 at 18:29 | history | answered | James Huddle | CC BY-SA 4.0 |