phc-discussions - Re: [PHC] pufferfish

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]

Date: Mon, 24 Mar 2014 12:39:49 -0400
From: Bill Cox <waywardgeek@...il.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] pufferfish

On Mon, Mar 24, 2014 at 12:13 PM, Solar Designer <solar@...nwall.com> wrote:
> Can we really go much higher than 4 KB while retaining bcrypt's access
> pattern (frequent, tiny accesses)?  Should we?  Why didn't we in 1990s -
> did no one think of it, were we trying to save RAM (yet UFC-crypt in
> glibc wasted 128 KB on expanded and doubled DES S-boxes, IIRC even
> before bcrypt was introduced), was it simply because Blowfish was
> designed that way and bcrypt built upon it, or did we not want to incur
> the slowdown and waste bandwidth(*) by exceeding L1 cache size while
> doing those tiny accesses?  I think it's some mix of these factors, and
> the last one is still relevant and very important now.

Bcrypt's L1 resident access pattern sounds like a great way to
compute-time harden a password hashing scheme, but when a company like
Adapteva finally ships large die, rather than the tiny parts it makes
through MPW runs today, wont Bcrypt be rather insecure?

I like the idea we'd discussed before of a Bcrypt inner loop that
repeats many times, doing mostly reads (because of write-through
caches), and a then one write pass, and moving to the next block.
That might let you tune it for just L1 cache, L1 + L2 cache, L1 + L3
cache, or L1 + external DRAM.

Bill