phc-discussions - Re: [PHC] multiply-hardening (Re: NoelKDF ready for submission)

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: <20140213155643.GA6212@openwall.com>
Date: Thu, 13 Feb 2014 19:56:43 +0400
From: Solar Designer <solar@...nwall.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] multiply-hardening (Re: NoelKDF ready for submission)

Bill,

On Thu, Feb 13, 2014 at 07:47:28PM +0400, Solar Designer wrote:
> In fact, once we add low-level parallelism we also need something very
> much like a crypto hash to be applied at least once or twice per block
> anyway, for data mixing between the parallel lanes

I don't know how exactly you were adding SIMD (as an experiment), but
please note that occasional data mixing between the lanes is a must,
since otherwise it becomes possible to compute the whole heavy part of
the KDF as a few sequential portions, one per lane, with the memory
needs reduced accordingly (it'll be number-of-lanes times smaller).
(This is similar to how scrypt's p>1 allows for computation with less
memory when the parallelism is not fully made use of as such.)

For the random access portion of computation, it is also important that
the random index depends on output of all lanes in the most recent block
computed.  Otherwise some of the lanes may finish computation later.

Alexander