Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 22 Mar 2015 00:40:48 +0300
From: Alexander Cherepanov <>
Subject: Re: Extend AFL to fuzz as you want

On 2015-03-19 06:13, magnum wrote:
> On 2015-03-19 02:21, Alexander Cherepanov wrote:
>> On 2015-03-18 15:35, Frank Dittrich wrote:
>>> On 03/18/2015 01:12 PM, Alexander Cherepanov wrote:
>>>> I think this is a general question to be discussed in john-dev, not
>>>> limited by the needs of fuzzing or security in general. Talking
>>>> specifically about fuzzing, when you want to fuzz functions behind the
>>>> valid() it's easier to patch this specific check out of valid() for now.
>>> But isn't the purpose of valid() to make sure all the other format
>>> methods only have to work with sane/sanitized input?
>>> Why should we care about segfaults etc. that would only occur after you
>>> removed some of the sanity checks in valid?
>> That's an interesting question. Short answer: we should care about it
>> because such a crash could be due to a genuine bug. But it very much
>> depends on a particular hash/valid()/get_salt()/etc.
> Knock yourselves out guys, just do not open GitHub issues for crashes
> that you can only trigger after removing whatever protection we had in
> place for them. Not unless you actually *did* find a genuine bug *and*
> can describe (or fix) it.

Sure, removing such safety checks is only a trick to make deeper fuzzing 
easier. After something interesting is found one has to check if it's 
applicable to the non-modified program. If a genuine bug is found 
usually it's easy to modify the sample to be acceptable the non-modified 

That was an easy part, now a bit more tricky one:-) There is a crash in 
7z format if you raise LINE_BUFFER_SIZE to, say, 1000000:

$ perl -le '$k = 300000; print q{$7z$0$1$$$$$0$}, $k, q{$0$}, q{1} x ($k 
* 2)' >
$ ./run/john
Loaded 1 password hash (7z, 7-Zip [SHA256 AES 32/64])
Will run 4 OpenMP threads
Note: This format may emit false positives, so it will keep trying even 
finding a possible candidate.
Press 'q' or Ctrl-C to abort, almost any other key for status
Segmentation fault

Is this a genuine bug in your opinion?

On a related note, I think it would be nice to drop most lengths from 
hashes. An attached patch both fixes the crash and ignores the data 
length recorded in the hash (as a first step in this direction).

Alexander Cherepanov

View attachment "0001-7z-compute-data-length-from-data-and-bound-it.patch" of type "text/x-patch" (1050 bytes)

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.