Talk:ISAAC (cipher)

This page should be combined with ISAAC as they are about the same thing. The ISAAC includes a link to Jenkins Web site on ISAAC which is the definitive reference.

I'll undertake the combination and redirection if no one else will, as time allows.

ww
 * Good idea. In fact the two articles are extremely similar! (ISAAC (cryptography) also includes a link to Bob Jenkin's website, BTW). However, I would suggest merging the ISAAC one with this one, rather than the other way around; there seems to be a common practice of naming ciphers with a (crptography) qualifier when the name can be interpreted as something else, which is often the case. Securiger 08:37, 26 Feb 2004 (UTC)

Bob Jenkins?
Is the link Bob Jenkins right? Is he really that Bob Jenkins? 87.217.57.201 15:23, 23 November 2006 (UTC)

I just looked at Bob Jenkins' (of ISAAC fame) website. He has a page with some sort of informal CV and it is pretty certain he is not the same Bob Jenkins written up in wikipedia. Actually "Bob Jenkins" has to be a fairly common name in English speaking countries. 203.164.223.20 10:11, 28 December 2006 (UTC)

This issue is obsolete. The link now appears to be to the correct Bob Jenkins. 125.255.32.22 (talk) 06:12, 21 August 2017 (UTC)

For the "citation needed" message, citation number 5 (the Aumasson paper) looks appropriate. Don't want to fix it myself, my editting skills are 10years out of date. 125.255.32.22 (talk) 06:16, 21 August 2017 (UTC)

ISAAC does not pass the TestU01 suite
"ISAAC passes the TestU01 empirical randomness test suite," is completely wrong. Not only is it not true, the reference given is to a primary source that very clearly doesn't understand how to run TestU01 tests. They ran birthday spacings and that's it. They clearly thought they were running at least the SmallCrush battery (which is only 15 tests), but they didn't. They ran one test and that was it. Even ignoring the primary reference problem, that is *not* "ISAAC passes the TestU01 empirical randomness test suite." That is "ISAAC passes at least one of 160 of the tests in the TestU01 empirical random test suite."

To be clear, I've done a ton of actual testing on ISAAC (taking countless hours of my CPU time), and it's very good empirically speaking, but that statement is completely and blatantly wrong. I'm a primary source though and so I will not link my own results. All good quality random number generators will fail one or two TestU01 tests sometimes depending on input. This is not just normal, but I would be very concerned about TestU01's quality if that weren't the fact. This is because true i.i.d. data absolutely should still result in test failures sometimes. What's far more important is that the failures are not consistent (they are not the same or mathematically related tests failing each time). Both ISAAC32 and ISSAC64 fail BigCrush in expected "random" ways (which is a good thing).

I've removed the statement and reference completely.

73.217.103.80 (talk) 23:42, 10 August 2023 (UTC)


 * Dear anonymous, the content you're referring to was properly sourced with sources matching Wikipedia policies.
 * On the other hand, your revert is based purely on your own personal opinion, in violation of Wikipedia principles.
 * If you're a knowledgeable expert, I strongly suggest you publish your claims on a validated platform or sufficiently credible blog.
 * In the meantime, I'll review your comments but I plan to re-add the content, as it's useful for the community. Wikipédien-franco-australien (talk) 02:12, 24 November 2023 (UTC)