|
Message-ID: <71d099023cfea55c9cbe386526c65ba0@smtp.hushmail.com> Date: Wed, 18 Nov 2015 23:52:22 +0100 From: magnum <john.magnum@...hmail.com> To: john-users@...ts.openwall.com Subject: Stacked modes In bleeding-jumbo, our support for "stacking modes" has matured and approaches stable & useful now. Not to mention understandable... We've always had rules, as a kind-of stacked mode upon wordlist (and single). And for a while, we have also supported rules with PRINCE. Then came mask mode, with it's "hybrid" alternative (which is just another word for stacked). So you can run eg. wordlist -> mask which will not only be way faster than rules, but can also end up GPU accelerated (with certain fast formats). For quite some while, this has been supported for "some" combinations and not very well tested. We've also had "hybrid regex" for a while, but for wordlist only. Over the last days, I've been adding, unifying, overhauling and testing all this and now we are supposedly at this point: * Mask mode can be run by itself, or stacked upon ANY other mode except single. * Regex mode can be run by itself, or stacked upon any other mode than single or mask. * We can even stack hybrid mask after hybrid regex. We always put mask mode last, since it may be GPU accelerated. * The chain can also end with an external filter in all cases, except a chain can not start with External mode, go via stacked regex or mask, and end with an external filter. Basically the chain of modes can start *or* end with External. BTW using external filter with GPU-side mask will end up in some more or less undefined behavior (we currently neither disallow it or warn). * Session/resume supposedly works fine with all combinations. Regex is worse than the others: In fact, regex mode can't save/resume at all by itself - but when running stacked, at least the parent mode [and possibly a downstream hybrid mask] will. For "large" regex's there might be quite some dupe work after a resume. If something can be done with regex *or* mask mode, you should always choose mask! So the "longest chain" of candidate generation we can currently do, is: $ ./john -wordlist -rules -regex:(...) -mask:(...) -external:(...) (or the same, but using -prince instead of -wordlist). Although the common case (for GPU acceleration) would probably be much simpler, eg. just adding -mask:?w?a?a to incremental or markov. doc/MODES and some other files are updated accordingly. magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.