phc-discussions - Improving Makwa and arguing points for or against

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-next>] [day] [month] [year] [list]

Date: Sun, 14 Sep 2014 05:13:11 -0500 (CDT)
From: Steve Thomas <steve@...tu.com>
To: "discussions@...sword-hashing.net" <discussions@...sword-hashing.net>
Subject: Improving Makwa and arguing points for or against

I think Makwa has one nice feature: delegation. The fixes needed are: required
pre-hashing, post hashing, and clearing p and q right after they are generated.
Replacing e=2**n with e=65537**n also require gcd(65537, (p-1)*(q-1))=1. This
way you can say it's as safe as RSA. I guess that's not really necessary since
post hashing is now required.

Fixing these problems removes the touted features password escrow (worst idea
ever), fast path, and offline work factor (Although this is not "true" since any
hashing algorithm can. With an upgrade parameter that just loops calls of the
main algorithm using the password as the hash from the previous one. I
understand that this upgrade is literally increases the work factor [ie Hash of
cost=10 then upgrade cost by 10 is the same as hash of cost=20]. This is needed
for delegation to work smoothly with upgraded hashes).

In respect to the time cost, if you know p and q it runs in constant time. This
is for fast path if you keep p and q you can do m**(2**x) (mod n) super fast but
an attacker needs to do x squarings. This is a server-specific shortcut and was
given has something interesting for submitters to have. I thought yeah that is a
cool idea, but when you compare it to "server-specific key", "server-specific
shortcut" is worthless.

Delegation is cool, but I don't see it as something that will be used in the
real world. The only arguing points I see are: is delegation awesome enough to
make it go to the next round, whether we can force changes to an algorithm, and
whether we can prevent the author from saying "you could keep p and q for fast
path".