[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20150822005455.GA6971@openwall.com>
Date: Sat, 22 Aug 2015 03:54:55 +0300
From: Solar Designer <solar@...nwall.com>
To: john-users@...ts.openwall.com
Subject: Re: Anyone looked at the Ashley Madison data yet?
On Fri, Aug 21, 2015 at 11:21:34AM -0500, JimF wrote:
> Here is a quick 'survey'. I simply took the first 265k hashes, and
> and started a forked check using just a few passwords.
To more reliably determine the ordering of the usual top passwords, I
recommend taking not "the first 265k hashes" (or whatever count), but
rather a random sample. You can do this by running "shuf" (part of
recent GNU coreutils) first, and only then using "head" on it.
> One thing I have also seen, is it may be best to do a -fork and
> OMP_NUM_THREADS=1 when using -single mode, since it keep the number of
> candidates and targets minimized. All work will be against only the hash
> that 'should' get the work done for it (up to a point). This hash is
> SO slow, that every option to try only the RICHEST set of candidates has
> to be done.
Of course. Better yet, "./configure --disable-openmp".
Related: your "just a few passwords" should be a multiple of
max_keys_per_crypt, which for bcrypt without OpenMP is either 2 or 3 on
a modern system (depending on your build). (With OpenMP, it is much
higher, and you correctly point out that this makes things worse.)
So if your wordlist has e.g. 5 lines, it is wasteful - you could as well
test 6 in the same time.
Alexander
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.
