phc-discussions - Re: [PHC] Argon2 improvement thread

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: <20150721193757.GA4345@openwall.com>
Date: Tue, 21 Jul 2015 21:37:57 +0200
From: Solar Designer <solar@...nwall.com>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Argon2 improvement thread

The Argon team -

Congratulations on winning the competition!

Personally, I attribute this success largely to you managing to strike
the right balance that the diverse PHC panel members could agree on
Argon2 as the (basis for) the winner.  That's great.

On Tue, Jul 21, 2015 at 07:39:16AM +0000, Jean-Philippe Aumasson wrote:
> Argon2 will be the basis for the final PHC winner. What should be change to
> make it better than it is now?
> 
> The designers proposed an optional update:
> * "smarter non-linear indexing (...) in order to flatten the memory usage
> over time"

I support this tweak, although I haven't reviewed it yet - I have only
read the relevant announcement in here.  In fact, Bill and I were
advocating making a change in this area.

> * BlaMka (from Lyra2) instead of BLAKE2b
> see
> http://permalink.gmane.org/gmane.comp.security.phc/3008
> https://github.com/khovratovich/Argon2/blob/master/Argon2.pdf (chap 3)
> 
> Solar Designer proposed to integrate MAXFORM in Argon2d.

I still do propose this.  In fact, whether the final Argon2 gains a
MAXFORM chain or not will most likely determine whether I recommend it
for password hashing use in libc crypt(3) and such or not.  As Argon2 is
currently defined (whether 1.0 or 1.2), it is not good enough as a
universal replacement for bcrypt, IMO.  I think it's very important that
we correct that.

Then, if/once Argon2 gains a MAXFORM chain, the switch to BlaMka may
possibly be reverted, for simplicity.  I have no strong opinion on this.
The MAXFORM chain can provide far greater multiply latency hardening.

This needs more work to arrive at a specific proposal, ensuring minimal
differences between the i and d flavors (ideally, we'd have a similar
chain in i as well, just without the password-dependent S-box lookups),
and overwrites of the (initially highly sensitive) S-box contents (one
idea I have is to reuse the existing intermediate write of 1 KB to
state, which is currently wasteful on systems with write-through caches -
at least we'd put it to use, increasing the amount of state that a TMTO
attacker would need to deal with).

Dmitry, Alex - what do you think on this?  How do we cooperate on it best?

> Bill Cox proposed (in his yesterday's email):
> "- A hybrid Argon2i/Argon2d (Argon2id?), where some initial fraction of
> memory hashing is done in a cache-timing independent manner, followed by
> unpredictable addressing to improve off-line attack resistance.  If this
> fraction were a parameter, it would unify Argon2i and Argon2d into one
> algorithm.

I also support unifying the two, with a parameter.  As to making it
tri-state (or beyond) rather than binary, I have no strong feelings.

> - Improved GPU resistance, similar to Yescrypt"

This fits under my MAXFORM chain proposal.

> The selected tweaks should make Argon2d and/or Argon2i better but without
> changing too much the original design (none of the above suggested changes
> would).

Alexander