Google recently filed a patent which shed some light into how they rank pages in searches. One of the observations the linked article makes is that google considers how long your domain has been registered for. I doubt I'm the only person to have thought "Hey, I wonder if that could be a useful rule for spam filtering?".

I have had some thoughts regarding spam filtering which are appropriate here. Since spam filtering is effectively an arms race between mail server admins and spammers, the important thing is not the tools but the method of making new tools. 'Meta-tools' perhaps. So, the thing that you need for effective, long-term spam fighting is a really good infrastructure for finding spammer patterns.

(all major innovations in spam filtering are a result of spammers having identifiable traits. From bayesian filtering, which looks at identifiable traits of the mails themselves, to graylisting, which relies on observing the way that spammers (ab)use the simple mail transfer protocol).

So in an ideal world, we'd have a mail-filter infrastructure where we could make an observation like above, write a quick rule, drop it into the set and see, statistically speaking, how useful that rule proves.


Comments