Re: automated spam detection

[Ragnar Lonn <]

On Tue, 23 Feb 1999, Clive D.W. Feather wrote:

> Carl Moberg said:
> >> When a message is spooled, an operator is alerted. Maybe by email. The
> >> filter asks the operator "Is this spam?" and the operator gets to reply.
> >  I think this has been said before...
> >   "breach of privacy"
> 
> and "unscalable".
> 

Yeah but responding to an ever increasing flow of complaints if
your users are spamming is equally unscalable. The system is only
useful if the work required by the admin to scan and OK mass-mailings 
is less than the work required to clean up after successful spammings
(by your own users or others). We have people dedicating a fair amount
of their time responding to spam complaints, even if it only means
sending a pre-written response, and if that work can be more or less
completely avoided, it would be worth a lot of time spent doing 
something else.

The serious problem with the method is, IMO, the one Carl mentioned -
the breach of privacy. 

It would be a lot better if the *receivers* had an easy way of generating
feedback, telling a mail server that a message is unwanted UCE. Would it
be possible to have the mail server insert an HTTP link somewhere in
the message?  Saying "click here if this message, in your opinion, is
spam"  How large percentage of all users today use web browsers to
read their mail?  Could an HTTP link be inserted in a message without
altering the message unacceptably?

A dedicated mail header could be used but mail clients would have to be
adopted to make use of it. That would be nice though - a mail header that
contained info for the end user mail client about where to report a
message as spam. Mail clients could have a large, red 'SPAM ALERT' button
that let users easily give feedback on spam they received - they would
simply click the button. It might be almost as efficient as having admin
people look at bulk mailings before OK:ing them for delivery, it would be
safer (as the receivers are the ones really capable of judging whether
something is unwanted/unsolicited or not), it would not constitute a
breach of privacy, and it would save more time for admins.

  /Ragnar