>> The idea being that you don't want any more content searching than is
necessary, particularly when a recipients-dictionary-attack is underway. <<

Okay, but if you wait until the message is stored in the queue and NOW you
have to scan each one with a command-line process - how is THAT better
(that's the transport sink scenario).

What you want to do is:

A) upon connection, check DNS BLs - if matches, add points

B) upon HELO, check HELO rules - if matches, add points

C) upon MAIL FROM - check for <>, if it matches, set a flag (there should
only be ONE recipient)
   check DNS BLs for blacklisted recipients, if matches, add points

D) upon RCPT TO - check for valid recipient - if more than 2 invalid
recipients, drop connection.
   If sender is <> and more than 1 recipient, drop connection
   If recipient is Postmaster@ or Abuse@ or Root@ (etc) and more than 1
recipient, drop connection (with proper return code "too many recipients)

E) at EOD (after the CR.CR), 
   - check for SMTP AUTH (so you can skip scanning)
   - otherwise scan the content with Sniffer (and Virus Scanner) - add
points
   
If the points exceed your threshold at ANY time, drop connection.  No bounce
message necessary, no need to store the message in the queue, etc.

Whenever you drop connection, add IP to your "tarpit/graylist" list.  Use
that for subsequent "upon connections"


>> Me, I like the idea of accruing a weight against the sending IP when a
recipient lookup fails.  <<

You can do that by processing the log file.



Best Regards
Andy Schmidt

Phone:  +1 201 934-3414 x20 (Business)
Fax:    +1 201 934-9206 



-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Colbeck, Andrew
Sent: Friday, February 18, 2005 08:06 PM
To: [email protected]
Subject: RE: Re[2]: [sniffer] IIS SMTP Integration


Pete, Matt was specifically referring to envelope rejection (as well as
other info gathering actions) based on address validation (and any other
criteria based on information as it can be tested, like a local blacklist
against the sending IP).

The idea being that you don't want any more content searching than is
necessary, particularly when a recipients-dictionary-attack is underway.

Me, I like the idea of accruing a weight against the sending IP when a
recipient lookup fails.  I get a lot of spam that is a single message which
is CC'ed and BCC'ed against a lot of addresses that are simply guessed, and
I want to punish those, and ideally, share that news with other mailservers.

Andrew 8)

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Pete McNeil
Sent: Friday, February 18, 2005 4:33 PM
To: Matt
Subject: Re[2]: [sniffer] IIS SMTP Integration


On Friday, February 18, 2005, 7:23:03 PM, Matt wrote:

M> Sanford Whiteman wrote:

>>Incidentally,  it  is  a  transport sink, not a protocol sink, meaning

>>that envelope rejection is not possible. I can't defend this as solely

>>a  choice  made for stability, as it was also a choice necessitated by

>>my  prototyping  in  VB (and, though it's been in production, it's not

>>much more than a prototype due to the lack of docs).
>>  
>>
M> Yes, that really is a key issue.  It needs to be a transport sink, or

M> at least work with one in order to prevent ongoing issues with brute
M> force spam floods.  I'm not sure that Peter from VamSoft understands 
M> the large market out there for non-Exchange based setups, or even for

M> going the extra mile that is necessary for this stuff, though that
M> might be an issue with resources and not just simply understanding.

Please give some more detail on this... When I researched this before it
seemed that a transport sink is good when you want the file, but if at all
possible you'd really want a protocol sink.

I had sketched out the idea of putting SNF up at the protocol level right
after CR.CR so that any mods could happen in RAM and so that if the message
were to be rejected it could be. SNF is up to the challenge because if you
can avoid all of the file system coordination stuff that the command line
version does you're down to periodically checking for a new rulebase file
and then running the scanner. That part of what SNF does usually happens
very, very fast. Faster, in fact, than most ping return times!!

If it could be done at that point in the process then why would you not want
to do it there? (Not a rhetorical question - I don't know enough about this
and want to learn.)

_M




This E-Mail came from the Message Sniffer mailing list. For information and
(un)subscription instructions go to
http://www.sortmonster.com/MessageSniffer/Help/Help.html

This E-Mail came from the Message Sniffer mailing list. For information and
(un)subscription instructions go to
http://www.sortmonster.com/MessageSniffer/Help/Help.html


This E-Mail came from the Message Sniffer mailing list. For information and 
(un)subscription instructions go to 
http://www.sortmonster.com/MessageSniffer/Help/Help.html

Reply via email to