Prevent brute force attacks on ejabberd server (with fail2ban for example)

Hi,

I recently installed an ejabberd server.
I would like to be able to prevent brute-force attacks on this server : it should not be possible to try thousands different passwords without being kicked out.

There already exists a tool that does this kind of checks : fail2ban ( http://www.fail2ban.org/wiki/index.php/Main_Page ).
This tool is able to monitor log files : if it finds too many authentication failures, the IP is temporarily banned (through iptables).
It works great on apache, ssh, etc. It also can be extended to other log formats (with regex matches).

I tried to make it work with ejabberd, but did not succeed because I faced 2 problems :
- fail2ban expects to have the IP of the client and the error message on the same line. But the ejabberd.log file creates two separate lines : an "Accepted connection", that gives the IP of the client, and a "Failed authentication", that is the monitored error message
- fail2ban expects to have a timestamp at the beginning of each line (see http://www.fail2ban.org/wiki/index.php/MANUAL_0_8#Filters ). ejabberd.log has the timestamp on a separate line

Do you think it would be possible to make ejabberd generate a log file suitable for fail2ban?
mod_logsession ( http://www.ejabberd.im/node/2999 ) looks very close to that feature, but it does not seem to log the IP of the client (see https://svn.process-one.net/ejabberd-modules/mod_logsession/trunk/README... ).
Another problem with mod_logsession is that it seems to require a recompile, which I would like to avoid, in order to benefit from the standard security upgrades of my distrib

Do you see another way to prevent brute force attacks? Maybe within ejabberd itself?

Thanks