Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Tue, 27 Sep 2011 08:29:16 +0200
From: magnum <rawsmooth@...dband.net>
To: john-dev@...ts.openwall.com
Subject: Some ideas about enhanced self-test

I take this on-list from here in case someone wants to chime in. This 
started as a private discussion about eg. crc-32 and dummy getting 
significantly inflated benchmark figures from including the null 
password in self-test. The opposite happens with eg. max-length 
self-test for NT - it would decrease the benchmark figure.

On 2011-09-26 02:21, JimF wrote:
> ----- Original Message -----
>  From: "magnum"<rawsmooth@...dband.net>
>> IIRC, the speed test is not necessarily made with all test hashes,
>> just some of them. We should establish just how this is happening.
>> Maybe dummy.c and crc32 could have the blank test in there, but
>> moved to a position where it won't be used for speed test? Testing
>> the null string is important for many formats (key cleaning issues
>> etc) so including them is a good habit.
>>
>> If this is not possible with current john core, maybe it should.
>> We should really have a 27 character test in NT_fmt but it would
>> currently make the benchmark show low.

> might not be bad to have a field, that tells how to use it in bench.
>
> 0 is test and bench all 'ways'.  (all defaults would have this).
> 1 is test and bench 1 hash
>  2 is test and bench multi
>  3 is test only.
>
> Or something like that.  Could be useful to then provide 'extreme'
> hashes for testing, while not skewing the benchmarks too much.  Bench
> should test 'normal' things, trying to get as close to an 'accurate'
> timing of the format.

I'm not sure I understand what you mean with '1 hash' vs 'multi'? 
Anyway, a couple more possible variants:

4 test-only and should match after truncation.
5 test-only and should NOT match the given plaintext. I don't remember 
now but I recall there are cases this would be relevant... maybe also 
related to truncation?
6 (might be overkill) plaintext in struct given in UTF-8 - bench.c 
should convert to current encoding, and if conversion fails, skip this 
entry.

Maybe it should be a bit field, eg. one bit corresponds to test-only, 
and so on.

magnum

Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ