spf-discuss
[Top] [All Lists]

"extreme SPF" scenario for ISPs

2004-01-30 18:24:33
Here is one possible "extreme SPF" scenario.  I do not expect the
average ISP to actually follow it to the letter, but it is useful for
keeping in mind as an ideal case.  In practice there is friction due to
old email clients, etc, but we can't let the past hold back the future.

1) ISPs are expected to be the point of control for outbound mail.
   Getting users to secure their machines is a lost cause.  ISP SMTP
   servers have to be responsible for performing virus and spam
   filtering.

2) ISPs ensure that all end-user accounts are funneled through the ISP's
   mail servers.  They can either use SS or just block port 25.
   Business DSL may or may not want to also take advantage of this
   protection.  This is already the recommended configuration anyway.

3) ISPs perform outbound virus filtering.  This is not a hard problem.

4) ISPs perform outbound spam filtering.  There are a number of
   approaches: you can do content filtering or rate limiting (develop
   usage profiles, throttle back if suddenly user suddenly sends 100 /
   minute) or some other approach or combination of approaches.  ISPs
   keep up with the spam vs antispam evolutionary race by upgrading
   their MTA pool.

5) ISPs are now responsible for blocking worms and viruses.  Many worms
   and viruses forge the return-path.  Let's restate the goal as "ISPs
   are now responsible for blocking forgeries" and see what that buys
   us.

   An ISP could require that outbound mail submitted through its servers
   must contain the ISP's return-path.

   This is a convenient way to prevent outbound spam and is not
   inconsistent with ISP "crackdown" policies.  Power users can work
   around by submitting elsewhere to 587.  Average users are using
   webmail anyway.

6) If the ISP already has SMTP AUTH enabled, you get a bonus: you can
   enforce that not just the domain but the entire return-path matches
   the local user, whose name is known thanks to AUTH.

7) If requiring a local envelope domain is too draconian, ISPs could
   waive the return-path restriction by performing *pre-emptive* SPF
   checks on *outbound* mail.  This solves the vanity domain problem;
   all a vanity domain would have to do is indicate the ISP in some
   form.  If the SPF test passes, the outbound mail would be accepted
   from the user and transmitted.  In the simple case this is compatible
   with (5).

8) The benefits of matching the return-path to the actual sending user
   are as follows:

   * bounces go to the right place; if a spam run manages to go out, the
     resulting bounces will be contained in the user's mailbox and not
     anybody else's.  As sender address verification becomes more
     popular, if a mailbox over quota the spam run becomes self-limiting
     because nobody will accept the mail.

   * you reduce liability (could envelope sender forgery be considered
     libel or a CAN-SPAM violation?  If a joe-jobbing virus/spam that
     passed through an ISP's mail server deliberately forges a target
     address rendering it unusable due to bounce volume, could that
     target sue the ISP for negligence?)

   * once you do this, you might not even need to put in outbound
     spam/virus content filters; the return-path is a very good
     approximation.

9) MUAs are encouraged to make the following changes:

   * display the "responsible sender" header (from Caller-ID) as well as
     the purported "From:"

   * allow configuration of envelope return-path / Sender header as
     separate from "From" address.

   * display the "From:" in full, not just the name part.  Phishing is a
     tough problem.  Consider the following scenario:

        From: service(_at_)paypal(_dot_)com (PayPal Customer Service)

     would be protected by some form of header authentication, but

        From: badguy(_at_)spammer(_dot_)net (service(_at_)paypa1(_dot_)com)

     meets or exceeds all technical requirements!

10) note: You can get around the unpalatable block-port-25 problem by
    doing outbound transparent SMTP proxying in much the same way as ISPs
    are already doing transparent HTTP proxying.  The argument stands.

-------
Sender Permitted From: http://spf.pobox.com/
Archives at http://archives.listbox.com/spf-discuss/current/
Latest draft at http://spf.pobox.com/draft-mengwong-spf-02.9.5.txt
Wiki: http://spfwiki.infinitepenguins.net/pmwiki.php/SenderPermittedFrom/
To unsubscribe, change your address, or temporarily deactivate your 
subscription, 
please go to 
http://v2.listbox.com/member/?listname(_at_)©#«Mo\¯HÝÜîU;±¤Ö¤Íµø?¡


<Prev in Thread] Current Thread [Next in Thread>