Hello,

My spam setup considered soon enough these possible SQL injection attack as 
spam.  I mean it's nice to know if someone may try something nasty. However, be 
warned 150 times about the same IP and in the end do nothing about it, that's 
just noise.

I considered modifying the script so it bans temporarily IPs of suspicious 
clients. I'm finally hesitating to do so:

- First, really, it's stored in /root/bin? Not trying to enforce LSB but 
creating a bin directory in /root is a bit wild, isn't? I known Vincent and 
Sylvain aren't happy at all with the idea of using dpkg/apt-get to maintain 
scripts and configs and I admit this method (old gnapgnap packages) is not 
flexible, a pain in the ass in some regards. But I'm not sure creating a bin 
directory in /root is the kind of flexibility that ease things. Easy to set up 
sure, but if I want to modify it without breaking anything, what should I do? 
How was it installed? Was it a standalone script? Is there any other version of 
this script around? Could I commit my changes somewhere? etc. Plenty of 
questions, just for one script.

- Second, a simple look at the script shows that scalabily was not took into 
account. With this script running, anyone could considerably flood all of our 
mailboxes and have the mail server running on this system sending thousand of 
mails in a way too short time span. 
When Gna! was created, Vincent had this nice script called logalert  
http://home.gna.org/logalert/  I'm not sure to understand why it was dropped 
but even if it required some time to configure it properly, seems to me a 
better starting point. For instance, I'm not sure it is so great to have a 
daemon to watch a single file instead of a cronjob. Granted, the daemon will 
catch the malicious log line when it appear, not 3 minutes laters. But since it 
only send mails about it, it may as well catch it 3 minutes later. And will we 
have a daemon for each new log file we may want to monitor?



Actually, I noticed that plenty of things disappeared over time and I do not 
know why; I guess I missed much anyway so it is not unexpected that I do not 
know why and when these are gone, I'm puzzled still, I understand these were 
maybe annoying in some ways but I'm getting afraid of a setup with many open 
doors and too few checks. If these were annoying, that's because somehow they 
addressed issues.
Simple crude example, I found no installed version of chkrootkit. Why so? 
Seeing that, I have the same feeling than years ago when I was poking poor Paul 
Fisher to get a kernel without the kmod/ptrace bug running on the original 
savannah.gnu.org box (first mail about it sent 27-03-2003, the upgrade occured 
16-06-2003... so we knowlegdly ran almost 3 month with a flawed kernel). We had 
no valid security policy, everything was flexible as hell, and when Vincent 
joined an had the _fun_ idea to run chkrootkit, we discovered we were 
compromised, remember?
I'm not saying the current setup isn't good. I just have the feeling that no 
one is able to tell exactly what it the policy on these machines about 
security. If so that would be mean there's a dangerous lack of vision on the 
issue. If not so, then I think this policy should be more clearly described 
(for instance it should be easy to know which security checks are made on each 
system and which scripts/programs should be installed on each system  - without 
such list, there's no telling if everything is right).
Other example, we have a file /root/TODO. Really? On Gna! we have to rely on 
text file somewhere on a computer? Not even a TODO with the Emacs intergrated 
feature, no a simple flat text. Does it mean Gna! web interface is not 
practical enough to manage tasks? If so, we ought to do something about it 
soon. Moreover, when people does not use the tools they provides, they quickly 
enough lose touch with theirs users needs, that's my pick on sourceforge. You 
cannot provide decents services that you make no use of.
Other example, /root/.ssh/ seems now managed by hand. There was a script that 
extracted ssh keys of Gna! relevant users.  If I create a new key, will I have 
to go over each system of put it or is there any other script that will do it 
for me? If why no longer using it? Was there a problem with it that could not 
be addressed by modifying the script? When adding a virtual user to access 
system to make backups, why doing it by hand instead of creating a virtual user 
on the web interface that we can easily track afterwards, add new keys, etc?

So I guess I'm missing info, overlooked stuff and am wrong on some or all 
points, but in the overall, that's a bit the feeling that comes to me that 
makes me unable to do anything. So tell me I'm wrong and how so and everything 
will be good :-)


(Don't take me wrong, I admire and welcome all the work done to get 
hardware/bandwidth up and running over years. I'm sure you understand that's 
not the point. The vserver-based setup is a fine one and obviously better than 
the chroot-based previous one. Everything is up and running and that's what 
matters most. 
I understand as well that, obviously, the set up should be practical in the 
eyes of the ones that maintains it most of the time. I'm just finding it 
however very hard now to contribute because I've not enough time to guess what 
policies could be, where stuff should be, etc)


-- 
Mathieu Roy

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
Project mailing list
[email protected]
https://mail.gna.org/listinfo/project

Reply via email to