Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Wed, 2 Jan 2013 22:52:42 +0100
From: Frank Dittrich <frank_dittrich@...mail.com>
To: john-users@...ts.openwall.com
Subject: Multiple formats accepting the same raw hashes

Hi all,

with the ever-increasing number of hash formats supported by John the
Ripper, it is inevitable, at least for raw hashes, that there are
multiple hash formats which "understand" the same raw input.
Yesterday, we had a short discussion of possible issues resulting from
this fact on john-dev, and we felt we might need input from john users
for the decision how to handle such cases.
(After a somewhat lengthy explanation, there will be a few questions for
you at the end of my mail.)

A recent example is the BLAKE2 format that has been added to the github
repository.

Prior to the introduction of this format, the following hashes would
have been interpreted as raw-sha512 hashes:
4245af08b46fbb290222ab8a68613621d92ce78577152d712467742417ebc1153668f1c9e1ec1e152a32a9c242dc686d175e087906377f0c483c5be2cb68953e
b109f3bbbc244eb82441917ed06d618b9008dd09b3befd1b5e07394c706a8bb980b1d7785e5976ec049b46df5f1326af5a2ea6d103fd07c95385ffab0cacbc86

$ ./john testpw
Loaded 2 password hashes with no different salts (Raw SHA-512 [32/32
OpenSSL])
guesses: 0  time: 0:00:00:02 0.00% (3)  c/s: 169771  trying: sposy - serrun
guesses: 0  time: 0:00:00:03 0.00% (3)  c/s: 195360  trying: hEd - borrix
guesses: 0  time: 0:00:00:04 0.00% (3)  c/s: 202503  trying: jazm21 - jethon
Session aborted


Now that BLAKE2 has been added to github, this is what happens if you
use the latest git snapshot:

$ ./john testpw
Warning: detected hash type "blake2-512", but the string is also
recognized as "raw-sha512"
Use the "--format=raw-sha512" option to force loading these as that type
instead
Loaded 2 password hashes with no different salts (BLAKE2 512 [32/32])
guesses: 0  time: 0:00:00:02 0.00% (3)  c/s: 139432  trying: critas -
monniel
guesses: 0  time: 0:00:00:03 0.00% (3)  c/s: 151740  trying: 49382609 -
49184819
Session aborted

This means, the above hashes are now ambiguous.
Because there was no --format= option on the command line, john happens
to use the first format which identifies one of the hashes in the input
file as valid, and uses this format (here: blake2-512).
However, john also mentions which other hash formats would support at
least one of the hashes in the input file (here: raw-sha512).

If you want to force a specific format to be used, you'll have to use
the --format= option, either if you just want to suppress the checks
which other formats support the hashes given in the input file, or if
you want to make sure a certain format gets used:

$ ./john testpw --format=blake2-512
Loaded 2 password hashes with no different salts (BLAKE2 512 [32/32])
guesses: 0  time: 0:00:00:02 0.00% (3)  c/s: 154519  trying: meneste - saa
Session aborted

or
$ ./john testpw --format=raw-sha512
Loaded 2 password hashes with no different salts (Raw SHA-512 [32/32
OpenSSL])
guesses: 0  time: 0:00:00:03 0.00% (3)  c/s: 152979  trying: crestert -
0700013
Session aborted

In most cases, formats which accept raw hashes allow to use a prefix, to
avoid ambiguity.

E.g., if --format= is not used, this hash will be interpreted as a
blake2-512 hash:
$BLAKE2$4245af08b46fbb290222ab8a68613621d92ce78577152d712467742417ebc1153668f1c9e1ec1e152a32a9c242dc686d175e087906377f0c483c5be2cb68953e

And this one as a raw-sha512 hash:
$SHA512$b109f3bbbc244eb82441917ed06d618b9008dd09b3befd1b5e07394c706a8bb980b1d7785e5976ec049b46df5f1326af5a2ea6d103fd07c95385ffab0cacbc86

When john finds the password, it always stores the format specific hash
representation in john.pot, not the raw hash, even if the input file
just contains raw hashes.

Also, when you start a john session without specifying --format, the
format which is used gets written into the .rec file.
That means, if you interrupt a session and later restore it, john will
continue to use the same format (even if you upgraded john, and the new
version knows more hash formats than the old version).


To summarize:
You don't risk to loose any work if a newer john version supports more
hash formats, but you can't rely on the assumption that future john
versions will per default use the same hash format as earlier john
versions used on the same input file.

To ensure that raw SHA512 hashes will only be interpreted as raw-sha512
format, we would have to implement solutions or find workarounds which
also have disadvantages.

We could not allow new formats to be added which accept raw hashes, if
the same raw hashes are already accepted by another format.
Disadvantage:
If you only have raw hashes, you might not know what hash algorithm has
been used to create these hashes. In this case, it is a feature that
john lists all the possible hash formats which you can try to use.

We could allow new formats to support raw hashes, but make sure that the
format which in older john versions is the only format supporting these
hashes will be the one which is used as the default format (i.e., will
be used unless --format= is specified).
Disadvantages:
Additional effort for developers when implementing new formats and for
future maintenance is required.
While currently the only format supporting a particular raw hash is
probably also the most frequently used format for such a raw hash, this
might change in future.
(IMO, this doesn't really apply in this case: neither raw SHA512 nor raw
BLAKE2-512 are good password hash algorithms, so I hope we won't see
many raw BLAKE2-512 hashes in the future. But we would aim to find a
general solution, nut just a crude workaround which only applies in this
particular case,)

So, at the end of a long mail, finally a few questions:

How important is it for you that john mentions which supported hash
algorithms might be used to crack a given set of hashes, instead of
silently using the first hash format which supported raw hashes of a
particular fixed length?

How important is it for you that the default hash format when starting a
new password crack session on the same input file remains the same even
across different john versions?
(The same question asked in another way: Would you like developers to
spend time on this "problem" instead of adding new hash formats,
improving performance of existing formats, fixing real bugs?)


Frank

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.