Making my own anti-spam filter

I'm wanting to move to using mutt for all my email accounts. However, one of my main stumbling blocks is figuring out how I will filter out spam.

I've tried SpamAssassin but the install was too difficult and then CPAN didn't have an uninstall option (?!) for a package I needed to uninstall.

So I'm thinking about trying to develop my own filters with procmail and whatever other tools I can figure out how to use.

Are there any general instructions about how spam filters work?

I know about checking the Received-from: headers and other headers, but I'm a little lost as to what to look for in them and how I would programmatically determine that a message is junk.

Your spam filter would surely blow.

Most filters like the ones in Outlook are tought to recognize spam from huge databases of spam mail. Some like SA even do lookups to internet databases to compare emails.

You should just take another shot at getting SA to work.

Look up Bayesian filtering. This is currently the best model for spam filters. It's a form of statistical pattern recognition; you mark each incoming mail as spam or ham and it learns to recognise the difference.

Paul Graham has done work in this area, has a number of essays worth reading: http://www.paulgraham.com/antispam.html

Or maybe you could just try Bogofilter. It's not as powerful as Spamassassin, but it's simpler to use.

lol bayesian filterng

get a clue, this guy can't even install spam assassin, he is not going to set up a bayesian filter from scratch