[Top] [All Lists]

Getmail: A New Email Delivery Architecture to Control Spam

2004-06-01 15:36:57

Apologize if it is not allowed to post paper here.

Attached please find an abtract paper of Getmail. Any comments are
appreciated. (below is some high level description of the architecture and
some of the advantages.)

We propose a new message delivery architecture, called Getmail.
The principal idea behind Getmail is to force the spammer to dedicate
storage for the Emails being sent out. Instead of the entire message
being directly delivered to the receiver (or more precisely
the receiver  side Mail Transfer Agent,  MTA) from the  sender, only the
message header (including the ubject line with a limited length) is
sent to the receiver. The message  header is augmented with a field informing
where the  receiver can retrieve  the complete Email.

There are several advantages of this architecture. First, Getmail forces
the senders to maintain their messages and keep their MTA servers up,
before receivers retrieve the messages. Thus Getmail makes it relatively
easy to
identify spammers and facilitates the  design of schemes that rely on IP
addresses to block  spam messages.  Of course, spammers  can still crack
into other users'
machines to send spam, but clearly  it is orthogonal to the problem we
are dealing  with here and beyond the scope  of this paper.  Second,  in
current  system it costs  money and  time for  receivers to  receive and
handle spams,  this is especially  true for users who dial  up to
their  ISPs.  By  only  delivering  message  headers  from  senders  to
receivers, less  bandwidth, storage,  and time will  be occupied  at the
receiver side, and
senders will have a greater share of responsibility in managing the
messages they send. For example, senders hold their messages until they
are retrieved by receivers and clean up the messages that are not read
by receivers for a certain amount of time.
Third, by forcing sender side
MTA servers to hold messages,  spammers will notice that a large portion
of  the  spam messages  are  not  read at  all.  As  a consequence,  the
incentive to send spams is reduced.  Moreover, in the
long run, spammers  will learn who are interested  in a certain specific
advertisement  (the ones  who  read the  spam  message), and  commercial
advertisement  can  be  targeted   at  a  certain  group  of  interested
receivers, instead of being sent blindly to everyone.  In this way, both
advertisers and Internet users will benefit from this learning process in
a long run.

Zhenhai Duan
Assistant Professor
Department of Computer Science
Florida State University
Tallahassee, FL 32306-4530

Phone: (850) 645-1561
Fax:   (850) 644-0058
Email: duan(_at_)cs(_dot_)fsu(_dot_)edu

Attachment: paper.pdf
Description: Adobe PDF document

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