The future of security

[bear]

On Sat, 29 May 2004, Peter Gutmann wrote:

>"Anton Stiglic" <astiglic at okiok.com> writes:
>
>>I think cryptography techniques can provide a partial solution to spam.
>
>No they won't.  All the ones I've seen are some variant on the "build a big
>wall around the Internet and only let the good guys in", which will never work
>because the Internet doesn't contain any definable inside and outside, only
>800 million Manchurian candidates waiting to activate.

I tend to agree with Mr. Stiglic.

Cryptographic techniques can provide a few partial solutions to spam.

What cryptography *can* do is limit the possible senders to a known
list.  This has positive, but limited, utility.  If there's a single,
general list that more than a few people all use, then spammers will
be on it (or at the very least people whose machines spammers use will
be on it) and the situation is generally unchanged.

If everybody maintains their own list of people whom they will accept
email from, then email becomes much less valuable because it's no longer
a way to reach anyone who hasn't put you on their "good senders" list
or hear from anyone whom you haven't put on your "good senders" list.

Another thing cryptography can do is make it much harder (perhaps even
impossible) to spoof mail headers.  Imagine, for example, a protocol
where your machine recieves a "can I mail you?" message from some machine
out there in untrusted space, responds by sending a unique password or
key to the address in the "can I mail you?" message, and then recieves
email using that password or key.  This ensures that every piece of spam
you get must correspond to a password or key that you know where you
sent.

However, this is also of limited utility.  It hasn't actually stopped
any spam; it's just fixed it so you know whence a message comes.  How
can you use that knowledge?

If you know where spam comes from, you can send a spambounce message
that names a particular machine.  It's probably not the spammer's
machine.  It's probably just a machine out there that was running
windows or something so the spammer took it over and is sending
email from it.  The owner of the machine has no knowledge whatsoever
that his machine is trying to email you.  What will your spambounce
mean?

Here's where it all breaks down.  In some cases, we've seen people
trying to claim they'll arrange it so spambounces cost the sender
money.  But here we get to repudiation of charges; if a thousand
spambounces cost fred a thousand dollars, and all he did was run
windows and connect his machine to the internet, fred's going to
fight the charges.  He may win.  And whatever happens at that point,
it's not going to be costing the spammer any money.

In other cases, we've seen ideas for fred to post a separate bond for
everyone he sends email to; the idea being that his "can I mail you?"
message contains the address of some bank somewhere that can be
checked for the existence of the appropriate bond before the "okay you
can mail me" response goes back.  The idea here is that if fred does
not actually want to mail you, then fred will not have put up money
for the privelege of mailing you, so you will simply reject his
request.  The problem here is twofold; first, it means you have to put
up some money (amount indeterminate) for every email address you send
mail to.  This doesn't fly real well in countries with a steep
currency exchange rate.  It stops a spammer who can't get into fred's
wallet from using fred's machine to send you spam, but invites the
usual suspects to develop "integrated" mail clients that will automate
the bond-posting, enabling the spammer to get into fred's wallet.  At
that point, email fraud has escalated to financial fraud, and fred is
the victim.  The spammer who is able to get fred's machine to post
bonds can clean out fred's wallet.

There are partial solutions.  Each has problems.  As Mr. Gutman
writes, it's a social problem and doesn't really admit purely
technical solutions.  What technology can do is shift the problem
around a little, and *try* to shift the problem onto the spammers -
but the successes are always partial and in some way unsatisfactory.

Spam won't stop until spam costs the spammers money.

			Bear


---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo at metzdowd.com