On Tue, 17 May 2005, Dennis Peterson wrote:
>
> How would you handle the PTR record for an SMTP server that hosts 500
> virtual domains?
> 

Yes, I realize that getting everyone to change would be a pain in the 
butt and if we can do the following it would certainly reduce spam.  We 
host many domains and I can't think of a reason that it would break our 
virtual domain system since rDNS(IP) == HELO == SMTP's 220.  This is not 
to say that a spammer can't put a system like this together, but if they 
do it will certainly be easier to blacklist.  This won't get rid of it 
all, but it should drop rouge virus mailers with their own smtp-sending 
engine.  IMO, a sending MTA should never have its smtp port closed unless 
it is an end-user.  If they are an end user then SASL should be used to 
authenticate.  Dynamic SMTP servers are ok provided that the constraints 
below are accurate.  

If you ignore SASL authenticated connections, we can better authenticate 
mail connections with the following list of constraints:

1.  fDNS(rDNS(IP)) == IP  # trivial
2.  rDNS(IP) == HELO      # should be trivial
3.  rDNS(IP) == IP:smtp's 220 string.  
4.  SMTP FROM domain has an MX # trivial
5.  SMTP FROM domain MX has a 220 string of itself, rDNS or HELO.

Caveats: (please add your caveats here)
#3 & #5: Sending server must have something on port 25 to issue a 220
string.  This server does not need to have any more than a 220 
response, though it should be friendly enough to wait for a quit.  This 
can be done with a few lines of perl.

We don't implement this 100% but our system is moving that direction.  We 
will also tie SPF to the list of constraints.  Those who send email 
through us as their mail gateway will use SASL.  

For what other reasons might this not work?  What can we do to fortify 
this?

-Eric
_______________________________________________
http://lurker.clamav.net/list/clamav-users.html

Reply via email to