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  PHC 
Open Source and information security mailing list archives
 
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Tue, 19 Feb 2013 09:54:54 -0800
From: "Dennis E. Hamilton" <dennis.hamilton@....org>
To: <discussions@...sword-hashing.net>
Subject: RE: [PHC] Required Output Key

The common approach to this (illustrated by PBKDF2 but also other key
derivation functions) is to ask for a wider result and break it up into the
cryptographic parameters that are needed.  This happens, for example, with
authenticated encryption where an IV, symmetric key, and MAC key are all
needed.

So it is not necessary to have PHS( ) produce multiple results, it can
produce a wider result than just what is needed as an authentication
verifier.  

However, a successful attack on the part used as a password authenticator is
likely to be a successful attack on the rest of the PHS result.  Once the
authentication hash is compromised, one must assume that the key derivation
is as well.  (The negligible prospect of a collision on only the
authentication hash is a tolerable cost of doing business for determined
adversaries.)

It seems to me that we're right back at the problem of disclosed hashes and
the prevalence of poor passwords in situations where an off-line attack can
be carried out.

 - Dennis


-----Original Message-----
From: havoc [mailto:havoc@...use.ca] 
Sent: Tuesday, February 19, 2013 06:36
To: discussions@...sword-hashing.net
Subject: [PHC] Required Output Key

It is important for the submissions to return two values:

(1) A standard hash for verification.
(2) A variable-length encryption key.

If an attacker has access to the hash list, they'll probably have access
to other data in the database (user's name, credit card number,
messages, etc.), and assuming a full compromise, the only way to secure
them is by encryption with a key derived from the user's password.

It's easy to create two different hashes of the single output and use
one as the hash and the other as the key. But if what's standardized
only defines a verification hash output, coders are going to have to
implement that themselves, and they're going to invent all sorts of
wacky and insecure stuff.

Some mistakes I can see happening:

1. Hashing the hash to get the key.
2. Using SHA256(password) as the key.
3. Re-running a different stretching algorithm on the password.
4. Using the hash as the key and a hash of the hash as the hash.
5. Deriving a 32-byte key from the 16-byte output.

... and so on ...

These are obvious mistakes to us, but we are cryptographers, and the
majority of coders using PHC are not. So either PHC should require
separate key/hash output parameters, or we should, in parallel, decide
upon the best scheme for deriving a (key,hash) pair from the single output.

What do you think?

Regards,
-- 
Havoc
(sorry for the nym -- I'll use my real name someday)

Powered by blists - more mailing lists