minimizing bandwidth taken be spam delivery attempts even when identifiable by "To:" address

["Daniel B." <REMOVEdanielTHIS@fgm.com>]

What anti-spam method minimizes the network bandwith used by spam
delivery attempts?

How hard is it to refuse incoming TCP connections to the SMTP port
based on DNSBL, using exim4?  Would refusing connections reduce the
overall traffic (maybe even causing spammer machines to think I no
longer run an SMTP server?)?  Or do spammer machines usually check
DNS MX records and would they deliver mail to the backup mail server,
which would then try to deliver it, using up the same bandwidth?

Is rejecting the message at the RCPT command the best action, or
is something else better?


NOTE:  My main problem here is not spam to valid e-mail addresses--
it is spam to invalid (non-existent) addresses.

Although it s easy enough to recognize the spam messagse and refuse
them, just the network traffic from the many spam delivery attempts
seems to use up a lot of my (modem-based) network bandwith.

Also, I've tried configuring DNSBL querying in exim4, but the extra
DNS packets seem to use lots more bandwidth.  (No, I haven't set up a
local DNS server/cache yet.)


ADDITIONAL NOTE:  This is it not for my main e-mail (which goes to
my ISP's SMTP server and stays there until I POP it down).

This is for a server I run on my own machine so I can define my own
e-mail aliases (which all forward back up to my ISP's SMTP server
and my ISP-based e-mail address) so I can avoid giving out my "real"
e-mail address to everyone and can detect who sells/rents/gives my
e-mail address (one of the aliases) to others.


Thanks.

Daniel