| 11:26.46 | *** join/#utah emcnabb (~emcnabb@hs-nc-1fff767f54-29977-1.tingfiber.com) |
| 11:45.23 | *** join/#utah kb_ (~kb@unaffiliated/kb) |
| 11:45.40 | *** join/#utah magmu_ (~Magical_M@174.27.155.81) |
| 11:47.38 | *** join/#utah tomeks (~tmks@turley.io) |
| 14:14.21 | *** join/#utah josephscott (~josephsco@74.114.153.251) |
| 14:14.59 | *** join/#utah josephscott (~josephsco@74.114.153.251) |
| 14:56.36 | *** join/#utah emcnabb (emcnabb@nat/redhat/x-syqbcuashwwaiigy) |
| 14:57.10 | *** join/#utah youtah (~youtah@67.139.187.132) |
| 14:57.10 | *** mode/#utah [+v youtah] by ChanServ |
| 15:07.06 | *** join/#utah youtah (~youtah@67.139.187.132) |
| 15:07.06 | *** mode/#utah [+v youtah] by ChanServ |
| 15:09.03 | *** join/#utah youtah (~youtah@67.139.187.132) |
| 15:09.03 | *** mode/#utah [+v youtah] by ChanServ |
| 16:23.56 | *** join/#utah braxtron (~braxtron@216.160.248.178) |
| 16:59.22 | *** join/#utah josephscott (~josephsco@74.114.153.251) |
| 17:03.01 | *** join/#utah tiwula (~lane@70.89.246.241-Busname-UT.hfc.comcastbusiness.net) |
| 19:59.49 | *** join/#utah emcnabb (~emcnabb@hs-nc-1fff767f54-29977-1.tingfiber.com) |
| 20:08.29 | *** join/#utah youtah (~youtah@204-15-84-228.static.ip.veracitynetworks.com) |
| 20:08.29 | *** mode/#utah [+v youtah] by ChanServ |
| 20:08.48 | youtah | ~cybertruck |
| 20:09.22 | youtah | infobot cybertruck is "breaks spangborn's window" |
| 20:09.22 | infobot | okay, youtah |
| 20:09.29 | youtah | ~cybertruck |
| 20:09.29 | infobot | somebody said cybertruck was "breaks spangborn's window" |
| 20:09.33 | youtah | oh dang |
| 20:09.34 | youtah | lol |
| 20:10.34 | youtah | ~cybertruck |
| 20:10.34 | infobot | breaks spangborns window |
| 20:10.37 | youtah | lol |
| 20:10.40 | youtah | perfect |
| 20:10.52 | jfindlay | lol |
| 20:13.33 | jlp_znc | https://www.dropbox.com/s/mup33obr3wf4r6s/red_ryder_anno.png?dl=0 |
| 20:16.54 | youtah | lol |
| 20:43.54 | spangborn | This snow is youtah's fault |
| 20:56.08 | jfindlay | eightyeight: do you think a secret key and its mapping into a smaller character set have the same entropy? |
| 20:56.36 | jfindlay | ent(sk) == ent(base64(sk))> |
| 20:56.42 | jfindlay | ent(sk) == ent(base64(sk))? |
| 20:58.04 | jfindlay | will the attacker have more fortune checking obscure chars vs checking longer keys of basic chars? |
| 20:58.47 | eightyeight | entropy is a measure of a process, not the result |
| 20:59.27 | jfindlay | so, it's like a thermodynamic property? |
| 21:00.29 | eightyeight | if `sk' was created by a CRNG picking elements from a set uniformly, then the entropy in the process measured in bits would be log2(set_size^numb_elem_picked) |
| 21:00.55 | jfindlay | but a process measure is irrelevant once the sk is generated |
| 21:01.00 | eightyeight | right |
| 21:01.11 | eightyeight | once `sk' is known, its "entropy" is zero |
| 21:02.29 | eightyeight | but hex(sk) = `BD7367DADD7FE017F58149D5CAC22238697ABCA9FD47B30B31ACB4092B1214DD' and base64(sk) = 'vXNn2t1/4Bf1gUnVysIiOGl6vKn9R7MLMay0CSsSFN0=' doesn't change anything |
| 21:03.04 | eightyeight | if `sk' came from a process with 256-bits of entropy, then how you encode it is strictly done for usability |
| 21:05.55 | eightyeight | alternatley, think of entropy as a measure of unpredictability |
| 21:06.19 | eightyeight | a mersenne-twister RNG has predictable output, once enough output has been monitored |
| 21:06.42 | eightyeight | so in that case, the RNG won't be entropic, as its output is predictable |
| 21:07.17 | eightyeight | whereas an RNG based on ChaCha20, seeded with a secret key, will produce output indistinguishable from true random white noise |
| 21:07.53 | eightyeight | no matter how much output has been monitored, no amount of analysis can predict the next bit (0|1) with any greater probability than 50% |
| 21:08.18 | eightyeight | so in that case, the RNG is fully entropic, as its output is unpredictable |
| 21:08.35 | eightyeight | does that make sense? |
| 21:10.35 | eightyeight | it's not wrong to talk about `sk' is its size. a 256-bit key means occupying 256-bits, and all 2^256 possibilities should be uniformly possible by the generator |
| 21:10.59 | eightyeight | so that's why we say "a key has n-bits entropy", even though it's probably better to talk about it in terms of its size |
| 21:30.18 | jfindlay | does ChaCha20(random_seed) really have a uniform probability distribution over that entire ensemble? |
| 21:32.55 | jfindlay | I was merely thinking about how password rules often require you to use 'special' characters to increase the eliteness of your pw |
| 21:33.36 | jfindlay | I guess it could be sensible to require that if you cannot increase the pw size |
| 21:34.41 | jfindlay | but with ascii, like half the available space is inaccessible because of the nonprintable chars |
| 21:41.26 | eightyeight | jfindlay: well, chacha20 keeps an internal state size of 512-bits, so if the set you're picking from isn't a factor of 2^512, then you're not picking your elements uniformly |
| 21:42.25 | jfindlay | what does it mean for a set to be a factor of 2^512? |
| 21:42.52 | jfindlay | do you mean the set of primes == factor(2^512)? |
| 21:42.59 | eightyeight | EG: if picking from the 94 graphical ASCII characters on a US/English keyboard, 2^512 mod 94 != 0, so you need to find an upper bound where that equation holds |
| 21:43.19 | eightyeight | otherwise, some outputs are more likley to be generated than the others |
| 21:44.26 | eightyeight | bring this into something you can chew on |
| 21:44.47 | eightyeight | suppose you want a random number between 1 and 10, but you are tossing a 12-sided die |
| 21:45.34 | eightyeight | you can't jus do 12 mod 10, as 1 and 2 are more likely to come out than 3-10 |
| 21:45.49 | eightyeight | so you should reject rolls of "11" and "12" |
| 21:46.23 | eightyeight | sucks you're throwing away 1 in 6 results, but this guarantees that 1-10 will be generated uniformly, provided the d12 is fair |
| 21:47.42 | jfindlay | what does a 2^512-sided die look like? |
| 21:48.00 | eightyeight | scale that up. if you have a 32-bit RNG, and you want to pick a random ASCII char, then you should adjust your acceptable results from the 32-bit RNG, to reject output that would introduce non-uniformity |
| 21:48.37 | eightyeight | seehttps://pthree.org/2018/06/13/why-the-multiply-and-floor-rng-method-is-biased/ for some examples in JavaScript how to handle that range |
| 21:49.21 | eightyeight | set your max number the 32-bit RNG can generate to be a multple of 94, and reject anything higher than that |
| 21:49.38 | eightyeight | then do your "roll mod 94" to get a uniform result |
| 21:58.33 | youtah | spangborn no |
| 22:02.08 | spangborn | youtah: yes |
| 22:02.21 | spangborn | You can't say no to deadpool |
| 22:03.01 | youtah | I did pretty well |
| 22:03.15 | youtah | No no - Consuela |
| 22:06.57 | spangborn | This is your fault youtah https://usercontent.irccloud-cdn.com/file/NAOL6cUF/IMG_1793.JPG |
| 23:15.51 | jfindlay | ~napoleon youtah |
| 23:15.52 | infobot | ACTION makes youtah feed Tina |
