1. Create a catch-all address for the domain you're going to use that isn't the normal postmaster one.
2. Pick a three- or four-letter combination of letters that rarely appears in normal conversation (like dcj, for example).
3. Set a mail filter on the catch-all account to forward all mail that has your three-letter combination as a part of its recipient list to your real address.
This means that for every service you sign up for, you can create a new address (I always use domainname.code@mydomain) that is trackable and gets to you. For example, I just signed up with Via Rail's online system - using the address viarail.dcj@mydomain. It will get forwarded to my real address (since it's got the dcj in there), if I start getting spam on it I'll know where they got the email address from, and if it gets really bad I can just change my filters to block all email to viarail.dcj@mydomain.
You could do something with Gmail's plus-addressing, but I find that many services don't accept those email addresses.
A free alternative is SneakEmail (http://www.sneakemail.com) which allows you to set up disposable addresses that forward to your primary account, and allows you to set up pre-forwarding filters. They also create a unique address for the sender of each email, and you can set up your SneakEmail filters to insert this as the reply-to address of each email you receive.
For 95% of the things I sign up for, however, I'm not that paranoid - the slight decrease in security is offset by the added convenience of not having to log into a third-party service (like SneakEmail) to get what I'm doing done.
Gmail lets you insert periods between the letters of your username. So if my email address is [email protected], the following are valid variants of my email address:
And so on. So you can use variants for different services you sign up for. Also note that you can substitute @googlemail.com for @gmail.com.
The only downside is that you have to manually verify your custom domain's 'sent from' addresses in Gmail, so you can't easily reply from arbitrary addresses.
My filtering now achieves around 99.6% filtering, and about 1 detected false positive per month. It would be interesting to see how gmail copes with my 2400 spam per day, and what accuracy it achieves, but it's a non-starter because of how I use email.