phc-discussions - Re: [PHC] Makwa is broken given p and q

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: <541122FD.4000200@ciphershed.org>
Date: Thu, 11 Sep 2014 00:20:13 -0400
From: Bill Cox <waywardgeek@...hershed.org>
To: discussions@...sword-hashing.net
Subject: Re: [PHC] Makwa is broken given p and q

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 09/10/2014 11:55 PM, Steve Thomas wrote:
> Given p and q you can do: e  = 2 ** cost e' = 2 ** cost (mod
> (p-1)*(q-1)) x ** e = x ** e' (mod p*q)
> 
> 
> You could pick the cost to be 2 ** 128. Without p and q you can't
> test a password. powConst = powm(2, pow(2, 128), (p-1)*(q-1)) hash
> = powm(password, powConst, p*q) but you could just do
> HMAC(password, secretKey)
> 
> Sorry but even if you came up with the perfect server-specific
> shortcut, HMAC or encryption with a secret key is better.
> 
> If you don't know the secret, it takes 3x longer. vs If you don't
> know the secret, you can't do anything.
> 

Hi, Steve.  I can see you've been trying to attack Makwa, just like
me, though probably with a better mathematical understanding to launch
such attacks.

Of course Makwa is broken if an attacker knows p or q.  It is pretty
worthless in that case.  Also, I agree that a master secret key is
better than a fast-path computation with p and q.  Delegation is the
thing that I like about Makwa.  The other features are not
particularly special, IMO.

Here's an "attack" slash "feature" I came up with today.  We can't
efficiently compute X^(2^w), but we can do so for g^(X^(2^w)) using
the binary decomposition of X.  This is true for all choices of
generator g.  In other words, you don't have to solve a particular
discreet log to break Makwa, but cracking *any* discreet log (with g a
group generator) will do.  I've been thinking particularly of powers
of 2 and it's inverse, but haven't gotten too far.

One cool thing -- in a mod n group, where n is odd, dividing or
multiplying by 2^m is approximately the same as doing an m-bit
bit-reversal.  When n is a power of 2, it becomes *exactly* a bit
reversal.

A possibly useful feature for Makwa making use of this would be
password specific fast-paths.  If we derived X, and then computed X' =
g^X, and used X' everywhere rather than X, then the server could save
the original X, encrypted with a master key, which would allow the
fast-path to be applied per password, reducing the likelyhood of a
disaster such as leaking p or q.

Or... we could simply have the password hash directly encrypted with
the master key.  That would be better.  And... better is gooder.

Bill
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
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=prun
-----END PGP SIGNATURE-----