A Plan for Spam describes how Bayesian filters be applied to spam e-mail (more technical paper on the subject is Native-Bayes vs. Rule Learning in Classification of Email.)  Pretty neat stuff even though statistics was one area I intentionally neglected during my physics days.  Microsoft was big on Bayesian a few years ago, so it makes me wonder why Outlook doesn't have a built-in Bayesian filter for spam instead of that annoying and undependable Rules Wizard.  I am tempted to write one myself, except companies like GiantCompany sells spam filters with similar result.  So I turned to the problem of blocking adult sites and pop-ups (pop-unders are no problem since Z-order of browser window is a good indicator.)

All one needs is a BHO (Browser-Helper Object) that applies Bayesian filter to web page contents.  If this is implemented already, let me know.