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: Tue, 05 May 2015 15:36:47 -0300
From: Marcos Simplicio <mjunior@...c.usp.br>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Argon2

On 05-May-15 14:54, Bill Cox wrote:
> On Tue, May 5, 2015 at 1:15 AM, Jean-Philippe Aumasson <
> jeanphilippe.aumasson@...il.com> wrote:
> 
>> FTR, the panel had agreed to accept Argon2 as a PHC candidate, superseding
>> Argon
>>
> 
> Awesome!  I do think this is the right decision.
> 
> As for potential tweaks, here's some dumb ideas:
> 
> - Argon2d/i are similar.  Can we make it one algorithm that has a flag
> specifying when to switch from cache-timing resistant mode to password
> dependent mode?  That would let one algorithm do both (something I can do
> in TwoCats)
> - Both Argon2d and Lyra2 are too focused on TMTO resistance, to the point
> of running slower than needed.  Cutting the Blake2 rounds in half in Argon
> considerably speeds it up, as does reducing the memory writes in Lyra2.
> This would improve both algorithm's basic defense.
> - Both Lyra2 and Argon2 should use the modified Blake2 round with
> multiplications for improved compute-time hardening
> 
> Argon2d is a late-comer and needs more work than Yescrypt or Lyra2, but it
> looks promising to me, if enough tweaks are allowed.  I don't know about
> the other finalists, but the two strong Argon2 competitors are Lyra2 and
> Yescrypt, both of which have already influenced the Argon2 design, I
> think.  I think both the Yescrypt and Lyra2 team would be happy to see the
> best possible solution, which could be Argon2 in the end, if the Lyra2 and
> Yescrypt authors were to cooperate with the improvements to Argon2.
> 
> I am a fan of the one-pass model in Argon2.  It could be enhanced into a
> winning algorithm, IMO.
> 

I still strongly believe that having a group of winners and a "chimera"
out of them would probably be the best outcome of the PHC, but the
future of the winner(s) is for the panel to decide :)

So, in the spirit of contributing: probably Argon2i would be better
using a Catena-like approach, i.e., instead of having a
password-independent pseudorandom visitation pattern, it would be better
to have a fixed and thoroughly analyzed visitation pattern. The
reasoning is (please correct me if I misunderstood something about the
algorithm):

1) For legitimate users, part of the PHS's computation time would be
wasted calculating the indices to be visited, while attackers could do
so only once and reuse the pre-computed indices in many threads,
diluting the corresponding costs. To avoid giving more advantage to
attackers than to legitimate users, the computation of indices should be
as lightweight as possible (which is a goal in Catena and also in
Lyra2's first pass).

2) Some salts/parameters will end up leading to weaker visitation
patterns than others. This is unlikely to be critical in the long run,
but that would be similar to having "weak keys" in cryptographic
algorithms. So, if those are avoidable, it would probably be better.

My two cents, at least.

Marcos.

Powered by blists - more mailing lists