On Feb 25, 2011, at 23:31 , Chris Woodbury wrote:
On Fri, Feb 25, 2011 at 1:53 PM, Patrik Karlsson <patrik () cqure net>
On Feb 25, 2011, at 00:15 , Chris Woodbury wrote:
Based on all of that, I have a few comments:
* Now that we have instance-name and instance-port, I'm not sure that
it's intuitive that the way to hit all instances is by specifying
instance-name=all, especially considering the case where there are instances
we don't know the names of. How about having a script argument like
I've added a instance-all argument to align better with the other names.
Does that make sense?
* I really like the idea that it's now possible to use SQL Server as a
proxy for doing Windows password testing. However, I think we should put up
some hurdles for the user. Virtually all the Windows networks I've worked
with have had account lockout policies in place, and I'm concerned that
someone could accidentally end up brute-forcing a domain and locking out a
bunch of accounts. I think we ought to add a script argument to enable it.
At first, I didn't understand the problem here, as you previously wrote
that you modified the script to abort once it detected a lockout.
When I tested it I noticed what happened, and that there's no way of
detecting that an account has been locked out based on the error message
returned by the server.
I agree that there's a risk of locking out a number of accounts,
potentially the whole domain, if you run the script without knowing what
your doing or how it works.
Either we document this with a few exclamation marks, or we simply fail
to start the script (if Windows authentication is used) and return an
explanation why and instructions how to proceed.
We could use an argument that tells the script how many attempts to
perform for each account, if unset and the authentication method is set to
Windows, the script won't start.
How about that?
Actually, since you mention it, it might be nice to have an "only do this
many attempts" argument for SQL Server authentication as well, since those
can also have lockouts. But maybe that's just something the user should
control with the password database (incidentally, the brute script now tries
the username as the password - should that be configurable?).
Anyway, I'd be fine with doing something like that for Windows at least,
or maybe just a ms-sql-brute.brute-windows-accounts argument, so that it's
explicit. I think either would be fine, as long as it's an extra step that
is required and is only applicable for Windows authentication.
I think controlling the amount of tries should be done by the number of
entries in the password database.
Since we detect account lockouts when sql authentication is used, we don't
need an additional hurdle here in my opinion.
I've made the following changes to the ms-sql-brute script:
* If the script is started ONLY with the mssql.domain argument, it returns
the following message:
| Windows authentication was enabled but the argument
| ms-sql-brute.brute-windows-accounts was not given. As there is
| way of detecting accounts being locked out when Windows authentication
| used, make sure that the amount entries in the password list
|_ (passdb argument) are at least 2 entries below the lockout threshold.
When the ms-sql-brute.brute-windows-accounts argument is given it runs as
Regarding changes to the ms-sql-brute script I would like to re-write it so
that it uses the brute library I wrote a while back. But let's merge this
Let me know what you think about all of that. In the meantime, I'm
going to do some documentation.
I took a shot at this in the attached patch; let me know what you think.
There's also a small bugfix in there in mssql.lua, where I changed a "host"
to "instance.host", which fixed a bug for broadcast-ms-sql-discover.nse
I've applied it but it failed on the ms-sql-discover script, so I did that
Also it removed the change to ms-sql-brute which makes it work together
with Windows authentication, so I added that again.
I've done a bunch of testing too and unless there's something else, I think
we should try to merge this?