Re: Other SMTP verification methods

Yakov Shafranovich <[email protected]> wrote:
> All of these callback schemes have possible DDOS and privacy problems, 
> not to say they should not be done but we must tread carefully.

  The callback schemes often involve SMTP-level callbacks.  This makes
them much more expensive to deploy from a protocol/bandwidth/time
point of view.  They're cheap to deploy from an administrative view.

  An identity publishing method is more expensive from an
administrative view, but cheaper from a protocol/etc. view.

> Another issue is that trust is not binary like in PGP - it probably have 
> many possible values (maybe 1 to 100, in PGP its 1 - 5?).

  Exactly.  We *already* know that a path of trust exists from my
domain to yours.  Sure, we don't know what it is, but we know one
exists.  So knowing the exact nature of the path doesn't help.  We
need to know how much confidence we have in that path.

  e.g. Path A results in me having a confidence level of 10%, path B
results in a confidence level of 90%, so I trust you to a certain
extent, and cache path B.

  The difficulty in this system is deciding how to create & manage
that confidence value.

> > [ MTA(IP, time at IP) database ]
>
> So it would simply keep stats on how long an MTA has been up?

  Sure.  Implementing it is fairly easy.  Maintain a DB of N domains,
and do periodic lookups, to see what the MX records are.  The only
cost is disk space & bandwidth.

  To make it viable, you'd probably have to charge a small fee to
maintain the record for a domain.  Have an automatic form for taking
credit cards & domain names, and the administrative costs become
small.

  There's a business in that somewhere.

> Would such system be vulnerable to a DDOS attack?

  Sure.  Nuke the main DNS server, and it goes down.

  But it could be distributed fairly easily.  Even with 10^7 domains,
the disk space would be cheap.  Fedex'ing a DVD of the latest database
every week to backup sites is fairly straightforward.

>  Would a spammer simply have a bunch of dormant domains?

  Sure.  But once they start spamming from them, they can quickly be
put in a blacklist, just like any other domain that spams.

  So to abuse this database, spammers would need to behave like
netizens for a long period of time.  As soon as they misbehaved,
another system could catch them.  Sounds good to me.

  Alan DeKok.