Re: How is it a solution to spam?

On Mon, May 31, 2004 at 12:31:02AM +0300, Stelian wrote:

> Each domain is used for a few hours for sending, making RBL very hard to do.
> After the notification hit the users inbox, you can't do much.

The one big selling point of IM2000 is that the sending machine has to
be traceable. For the recipients to be able fetch the mail from it, it
has to have a static IP or at least an MX record (or equivalent)

The system I'd always envisaged was a democratic blacklist.. if you open
a message and it's spam, you click a button which notifies your
blacklisting service. Once a critical mass of users have blacklisted the
same server, the subscribers' software automatically refuses to fetch
from there. You could make it hierarchical, so if the density of
spammers in a particular domain or netblock exceeds a certain threshold,
the whole thing gets blocked.

The damage done by false positives is mitigated, because they'll remain
on the sending server and can be resent when the server is removed from
the blacklist.

Finally, it gives the big webmail providers a much easier way of
stopping spam runs while they're still in progress. If they get a dozen
complaints about mail from a particular user, they can delete that
user's outgoing messages, potentially preventing millions of spams
getting out.. The spammers can keep making new accounts, of course, but
the sophisticated techniques that now exist for preventing automated
account creation should slow them down significantly.

The system doesn't have to be 100% perfect, it just has to make life
sufficiently difficult for the scum for them to find something else to
do with their time..

> At the same time, the bandwidth abuse of the servers remains constant,
> because it is limited only by the capacity of the spammers to send
> spam, and that remains the same.

Rate limit notifications from servers? If you're running a webmail
company or corporate mail server, it should be easy to spot anomalous
big wodges of notifications. You have to be careful about mailing lists,
but you can monitor which servers your users regularly poll, and
whitelist them.

It requires some degree of software sophistication, but it doesn't seem
any more complex than existing anti-spam solutions; I think it could
be effective.

> Although I like the design of the system, I think it makes sending mass
> emails more efficient, because it scales so good. Instead of repeating the
> message to each user, not knowing if they read it or not, you keep a single
> copy on your machine, and those who are interested come and download it.

So then maybe the spammers only spam the users that like spam? Seems
like poetic justice to me :)

> Goodbye webbugs !

I imagine many client implementations will prefetch messages they have
received notifications for, before the user opens their mailbox. So
while the fact that a message has been retrieves indicates it's been
delivered ok, it doesn't prove it's been read.

I can even imagine a system layered over IM2000 by which users can
specify that mail notifications for them are forwarded through a 3rd
party anonymizing service, so their IP address etc. are not revealed.

-- 
Home+FOAF: http://www.loth.org.uk/ * PGP: 201A57B6 * Original portions
© 2004 Steve Dodd  *  "I'm not falling apart, I'm coming together."  *
Appreciated this message? - http://www.loth.org.uk/tipjar/  *  TINC  *

"You may be sure, dear Crito, that inaccurate language is not only in
itself a mistake: it implants evil in men's souls."
	-- Plato