[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 13 Sep 2015 20:28:08 -0700
From: Fred Wang <waffle.contest@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: Re: Judy array
On Sep 13, 2015, at 7:45 PM, Sayantan Datta <std2048@...il.com> wrote:
> Nice!! Unlike perfect hash tables, Judy array are supposed to be cache friendly. However, I'm curious regarding the number of lookups required!!. I'll study them in more details.
>
> Fred, have you compared the performance of bloom filters vs bitmaps(maybe one or multiple)?
>
Yes. For the most part (and in particular, when Bloom filters are very sparse), they are always a win for "our" type of lookup. In cracking, the vast majority of hashes will fail, and that is what I optimized for.
A Judy array on its own is already faster than what John is doing. Fronting this with a Bloom filter vastly improves performance. Regretfully, I did not keep my bitmap timing - but it was not impressive.
I use a 10-year-old Dell 2950 as my test environment, precisely because it uses slower memory, and more easily shows improvements. For my "standard" test case (MD5, 29 million hashes, a ~13 million entry dictionary, and best64 rules, yielding about 1 billion hash attempts to find about 1.7 million solutions)
hashcat 3 minute 54 seconds
mdxfind 1 minute 15 seconds (Judy only)
mdxfind 47 seconds (Current code, Bloom filter + Judy)
Its important to note that this includes the time to read 29M hashes, and store them - this takes about 22 seconds on the test box. The box uses dual E5410 @ 2.33GHz, so 8 cores.
So, Judy on its own is pretty darn good, but fronting it with a Bloom filter is pure win for this application.
Please take some time to try it out.
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.
