Re: Gawker hacked: Another trove of password data

[Brandon Enright <bmenrigh () ucsd edu>]

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

On Sun, 12 Dec 2010 17:07:15 -0800 or thereabouts Fyodor
<fyodor () insecure org> wrote:

> It looks like Gawker (mostly a network of gossip sites) has been
> compromised.  The attackers posted more than a million usernames,
> email addresses, and password hashes:
>
> http://yro.slashdot.org/story/10/12/12/2234252/Gawker-Source-Code-and-Databases-Compromised
>
> That is obviously unfortunate for Gawker and their users, but it does
> give us more real-life password frequency data to use for improving
> Nmap.  It looks like the torrent file contains 1.2 million records,
> most of which include password hashes (some small percentage just say
> "NULL").  It looks like they are probably using crypt(), but I'm not
> certain.  The readme.txt says it is DES based and only allows up to 8
> characters, and the hashes are 13 chars long, so it seems like
> crypt().
>
> The torrent also includes cracked passwords for a subset of those DB
> records (188,281 accounts).
>
> I can easily add the 188,000 already-cracked accounts to the Nmap
> password frequency files, but does anyone have time and computing
> resources to start on cracking the rest?  I recall that Brandon was
> able to crack a very large percentage of the PHPBB password hashes we
> found before.  And I recall that members of this list scored very well
> in the Defcon password cracking contest this year :).
>
> Cheers,
> Fyodor

Thanks for pointing this out, more passwords is always a good thing!

I looked at this dump some earlier in the evening and I came to the
same crypt() conclusion as you, although I didn't actually try cracking
any of them that looked like crypt.

The other hash being used is bcrypt() (the ones that start with "$2a$")
which is an extremely well designed password hashing algorithm.

Unfortunately for us, both of these hashes are salted and pretty slow.
bcrypt() is so slow it makes cracking and exercise in futility.  I
don't think we will be able to crack a big enough percentage of them to
use them as a source of statistics.

Unless we get to say, 66% accounts cracked, I don't think we have good
enough stats about the passwords to add them.  If we can only crack the
easy passwords, then adding them to our stats will bias our data towards
only the very easily crackable passwords.  This could harm our existing
data.

If I come across something interesting/useful I'll be sure to report it.

Brandon
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.16 (GNU/Linux)

iEYEARECAAYFAk0FrnoACgkQqaGPzAsl94JbdQCgnMeh1tOtAgltCkV9aRpbGVCc
tFUAn2b5W3t6zYsF/3Oe6HISnoHN0QmM
=GHsR
-----END PGP SIGNATURE-----
_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/