phc-discussions - Re: [PHC] Argon2 modulo division

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: Wed, 22 Apr 2015 18:48:31 +0200
From: Krisztián Pintér <pinterkr@...il.com>
To: "discussions@...sword-hashing.net" <discussions@...sword-hashing.net>
Subject: Re: [PHC] Argon2 modulo division

Bill Cox (at Wednesday, April 22, 2015, 6:13:21 PM):

>             A.block_absorb(pwd, 16, pwd_len);

> This code runs in time directly proportional to the password
> length. 

it is the same for all password hashing schemes, and not hard to fix.
in fact, i already hinted at the fix in one of my emails here, in
which i called for the practice of not allocating any memory, and not
keeping secrets in the stack either, if at all possible. for example
gambit could look like

void gambit_128(char* entropy, char* mem, int t_cost, int m_cost,
  char* dkid, char* key, int keylength)

entropy would be the pre-assembled password-salt-pad101 buffer with
the size of a full keccak state (200), and the actual keccak algorithm
would work on that, not a local copy. presumably in place. mem would
be the memory buffer of size m_cost*8. so we avoid allocating heap
inside.

obviously, this just shifts the responsibility to the caller. but the
caller has responsibility anyway, since the password probably gets
copied multiple times between buffers on its way from the GUI to the
hash function. i have no clue how to fix that.

in fact, i feel i need to go and change the implementation this way.
gambit might not be in this competition anymore, but it is still in
the bigger PHC, aka life.

this is absolutely not an excuse for having other side channel
vulnerabilities.

> By the way, looking back at the Gambit code just now was a
> pleasure.  As much as we differ on defense philosophy, I have to
> credit you for writing clean code.  It's very nice work.

thank you.