lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CALn24pCTYGjakQXeVEh_J37KBn340YJBw5sJJanjq-_Y6Gg8qw@mail.gmail.com>
Date: Mon, 12 Aug 2013 10:32:33 -0700
From: "Bitweasil ." <bitweasil@...ptohaze.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] The EARWORM password hash

I really like the use of the AES intrinsic operations for performance.
 Unfortunately, right now, this does seem to make the algorithm very
x86-centric - I am not aware of AES hardware acceleration on ARM
processors, which means that this wouldn't work well for anything not-x86.
 However, it also (right now) badly cripples GPUs as they have to AES in
software.  I'm not sure that it would offer much of a benefit over FPGAs,
though if you've forced an attacker to FPGAs to get major speedups, you're
doing well.

I'm not entirely convinced of the merits of bandwidth limited access.  The
access patterns are certainly optimized, but on a busy multicore server
with, perhaps, 16 hardware threads, I don't know that you'd have great
streaming bandwidth to main memory.  The variety of accesses from all the
other threads may badly hurt the bandwidth delivered.  The attacker can run
on a quiet system for optimum memory bandwidth, and GPUs have plenty of
streaming bandwidth (though as noted, no AES primatives).

What does performance look like if you're doing a parallel build on half
your cores?  That should stress the memory system significantly without
choking CPU throughput too badly, and would resemble the type of activity a
busy webserver would likely see.  It may be the case that the memory access
to compute ratio is such that there is no penalty for this - worth trying,
though.


On Mon, Aug 12, 2013 at 8:57 AM, CodesInChaos <codesinchaos@...il.com>wrote:

> With memory hard schemes like scrypt it's easy to put a lower bound on the
> cost specialized hardware incurs per password guess.
>
> With bandwidth based schemes this isn't so obvious. Are there any papers
> analyzing this cost?
>
>

Content of type "text/html" skipped

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ