This is a general query to users of TMDA or other challenge-response
systems, just to get a sense of whether the end-users think the following
idea is a good one or a bad one.

If a given sender is well-identified and is known to have a good anti-spam
policy in force, should mail from that server be exempt from
challenge-response and delivered straight to the end-user? Would you
configure your system that way?  Should software (like TMDA) be configured
that way by default?

This is a little vague (on purpose) because I'm trying to find out the
opinion of the idea in general, rather than just one implementation, but
I'll give a couple of examples.

The company IronPort Systems (http://www.ironport.com/) is beta-testing the
system they call Bonded Sender(TM).  See http://www.bondedsender.com/ and
http://www.bondedsender.org/, where organizations put up money up front that
will get debited for any verified complaints (there is a dispute resolution
process).  This seems a good idea for businesses (at least medium-sized and
larger), although it may be too pricey for small-businesses, individuals,
non-profits, third-world businesses, etc.

I'm not connected with IronPort in any way, by the way.  My employer may
consider participating if there is enough utility.  For example, it would be
great if this really got us past all anti-spam software, including
challenge-response systems.  (I should emphasize that we DO NOT send spam;
we send things like invoices and lab results to users who have specifically
signed up for them.)  I realize that challenge-response systems have
mechanisms for pre-authorizing senders, but I think that many users will
find these too complicated and that the error rate will be high.

The implementation of BondedSender(TM) uses the DNS system.  Specifically,
you would look up the SMTP sender (just the last one before it enters your
domain) in a domain controlled by IronPort.  E.g. to check if SMTP server
192.168.1.2 was a Bonded Sender, you would query for A records for
2.1.169.192.query.bondedsender.org.

(Obviously this would leave dynamic IP addresses out.)

Besides this specific system, other implementations are possible, e.g.
cryptographic.  An SMTP-over-TLS server can be identified by its
certificate, although that by itself doesn't guarantee any given anti-spam
policy.  Also, individual messages can be digitally-signed, which is
advantageous for other reasons.

The issue of what kind of certificate to accept is very important.  I'm told
that many SMTP-over-TLS sessions use self-signed certificates, which do
provide some protection against eavesdropping, but don't provide any
guarantee of identity.  There are also some CAs that provide free
certificates for encrypting and signing email, but only certify that the
certificate holder had a given email address (at the time the certificate
was issued).

The advantages I see to exempting "certified" senders are this:
  1) automated mailings from certified senders will go through without
special intervention required.
  2) individuals would need to respond to fewer challenges and have fewer
valid messages in their pending queue
  3) could reduce the challenge-to-forged-from-address problem (*).


I think that it is possible that challenge-response systems could act as a
bridge between unauthenticated SMTP systems (currently all), and the brave,
new world of authenticated email.  The fact that people find
challenge-response systems slighlty annoying, and some people (I won't name
names) find them very annoying, could actually act as a stimulus to people
going through the hassle and expense of authenticating.

So, what do you guys think?  Would you consider exempting (e.g.)
BondedSenders?  What about messages that merely have identified senders, but
aren't linked to an anti-spam policy?

Ken Hirsch

(*) I think I should explain my thinking on item (3).  If a message comes
with a "From" address that indicates it is from a trusted domain, but the IP
address of the SMTP server indicates otherwise, then a challenge message
should not be sent and the message discarded.  Ideally, there should be a
reject (55x) message during the SMTP session with informative text (but not
a challenge).  I realize that most challenge-response systems currently act
after the SMTP session has terminated, in which case there is danger of
losing a valid message.

_____________________________________________
tmda-users mailing list ([EMAIL PROTECTED])
http://tmda.net/lists/listinfo/tmda-users

Reply via email to