On Tue, 2007-03-13 at 16:43 -0800, Paul G. Allen wrote:

> > 
> 
> Turns out we are black listed. The IP that is listed is 216.70.148.112.
> That's not our mail server either. In the past day I've found several
> things contributing to this problem. Once resolved I'll try to remember
> to post what I can here so others may learn from our mistakes (and learn
> why your IP(s) may get listed even though they shouldn't).
> 

OK, here's the end of the story.

There were two causes of the black listing, and two causes of our IP
being re-added to the lists after having been removed. The causes of the
black listing can come later. The re-listing is the important part of
showing where a mis-configured network can bite you in the butt. Kids,
don't do this at home (or at your company). Most of you on this list
probably do things the right way, but those who don't may learn from
this. BTW, I only became involved in this because our IT guy is on a
trip this week and the proverbial crap only hit the fan right after he
left. I am filling in for him this week.


Problem 1.

Our mail server (Exchange) has an external IP address, different from
our firewall. It is NAT'd through the firewall, and this should be a
Good Thing. However, when the mail server sends email, the outgoing IP
is not NAT'd to its IP, and instead uses the firewall IP. This is
*wrong* *wrong* *wrong*. Why?

When a slew of e-mails were sent from a user's computer (can you say
trojan?!), they were sent using the trojan's SMTP server and sent
through the firewall *using the firewall IP*. The result is that instead
of our firewall IP being listed, with no ill effect to the mail server,
essentially the firewall *and* mail server were listed.

When mail is sent through the one and only legitimate corporate mail
server, it should use its own IP address. That way if any other machine
on the network does a Bad Thing, the Bad Thing does not appear to be
coming from the mail server.


Problem 2.

Not filtering or blocking port 25 on the firewall so that nothing other
than those servers allowed to send mail, can send e-mail. Need I say
more?


Problem 3.

The PTR records and MX records on our ISP DNS servers are not configured
to support reverse DNS. Usually not a big deal, but on occasion there's
a system out there that wants to do a reverse lookup to see if the
computer sending the e-mail (or otherwise requesting a connection) is
who it says it is. In such a case, if the e-mail Received: portion of
the header does not match the reverse DNS, then the mail is rejected.
Reverse DNS in some cases must be requested of the ISP or they won't
provide it.


Problem 4.

Our mail server's HELO response was "quakeglobal.com" when it should
have been "mail.quakeglobal.com". This posed a big problem when trying
to de-list our IP. The CBL list wants to see a FQDN (Fully Qualified
Domain Name) as the name given in the HELO response from the mail
server. If it does not, it will re-list the IP. Because several other
lists use CBL as a reference, our IP remained listed on all of them
until the mail server response was corrected. Because several spam
filters use different lists, many people/companies could not receive our
e-mails.


Problem 5.

Our ISP did not update the ARIN records to indicate that our IP block
was ours. It was listed as belonging to a previous customer of the ISP.
So, doing a whois would come up with conflicting results. Even though
the ISP information was correct in all cases, going directly to ARIN
gave the wrong organization and contact address. Looking just now the
record still has not been updated, but they told us they would update it
ASAP.


The source of the listings.

Now, for how we got listed to begin with. Two reasons. First another
server other than our mail server sent a bunch of e-mails as part of an
engineering test. It sent too many too fast (engineering mistake). As
outlined above, these e-mails had the same source address as our
corporate mail server. When it was black listed, so was our corporate
mail.

The second problem was a trojan that infected a new computer. The trojan
was sending spam about buying pills online. Since port 25 was not
filtered, and the IP was again the same as the mail server,... well you
get it.


The Summary.

So, some very basic oversights and mistakes caused several days of
corporate wide problems. Some things that aren't usually a big deal
(like reverse DNS, a HELO response that's not a FQDN) only need to rear
their head once to cause you as an admin and your company as a whole a
ton of trouble.

I won't even get in to the anti-spyware/virus issues and the use of
Outhouse.

PGA


-- 
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Reply via email to