Re: Detecting brute force attacks

"Greg Metcalfe" <[email protected]> Wed, 25 Oct 2006 15:25:14 -0700
Newsgroups gmane.comp.security.linux
Message-ID <[email protected]>
On Tuesday 24 October 2006 13:41, Nic Stevens wrote:
> I wrote a script to detect brute force attacks by watching log files.
>
> If you want to try it it is here:
> http://www.ducksfeet.com/nic2/secwatch/index.html
This is a PHP script. I'm not sure that running PHP on the same machine as an 
Oracle server is the best of all possible plans. There's a rather large 
history of problems with PHP, and once it's installed people tend to use it, 
as nature takes it's course.

My take is that if you're going to do this via a log watcher* written in an 
interpreter, you'd be better off with something closer to a native package. 
As per the OP, that would tend more toward Perl, as it's installed on both RH 
and SuSE minimal systems. Or, if things are leaning more toward RH, it might 
make more sense to go with Python, as a RH environment tends to build those 
skills--closer to RH native (GUI-ack!) tools

OTOH, the OP hasn't been abundantly clear on terminology. In the circles I 
hang out in, there's been some discussion re: whether dictionary == 
brute-force attacks. Commonly assumed to be equivalent, yes. Validity, at 
least somewhat questionable. But that could all be written off to semantics.

I think we need more word from the OP. I'm not certain we'll get it, as so 
far, things are looking like, "I are a random Oracle CSSP, cannot adequately 
describe my problem, but nonetheless need help." 

Does a firewall or sshd logging solution meet the need, or does he require 
analysis of patterns (if any) in the accounts attacked and the passwords 
submitted? If it's the latter, then my post from a couple of days ago 
applies.

*If* the solution is yet another log analyzer, I'd recommend that before 
anyone rolls their own, they run off (quick like a bunny) and read Marcus 
Ranum's thoughts on how to do this.

Right now, we've no idea of what the OP needs, as he's made exactly one vague 
post, five days ago. I truly hope that this doesn't represent the internal 
state of the Oracle security group.

-- 
Greg Metcalfe