First of all, if you want to get rid of spam - configure your MTA correctly.
I use postfix, so here's a few links to a digested version of how UCE in postfix works, by Meng Weng Wong.
And here's the original lowdown on how postfix UCE works.
I'm working on adding information on SpamAssassin and Bayesian filtering, and a few practical tips on how I got my Postfix to repell spam - without (!) repelling one single piece of legitimate Email.
I will also tell you how I monitor what postfix actually does - via pflogsumm.
MTA Settings
I use this in my postfix conf (ie. /etc/postfix/main.cf):
# reject_unknown_sender_domain = give a 450 (rejected - try again later) to mailers sending mail with a from addr.
# that does not resolve to an A or MX record.
# The reason they are not just 550'ed is that it could be people with temporary dns problems.
# reject_non_fqdn_sender = reject mailers using MAIL FROM: unqualified adresses (such as localhost - or other which is definetely uncorrect).
smtpd_sender_restriction = reject_unknown_sender_domain, reject_non_fqdn_sender
Amavisd-new Settings
I use Amavisd-new to control my Spam and Antivirus measures.
SpamAssassin Settings
I use SpamAssassin to do my Spam Filtering.
NB. You need version 2.52 to do proper Bayesian Filtering.
For now I'll just include (md5: 03c40ce1fbc59da29dace2d99540b8a2) my Bayesian Database, so you don't have to do much teaching of your own :)
Ressources
Filtering E-Mail with Postfix and Procmail
how UCE in postfix works, by Meng Weng Wong