Spam
Well, I’ve been finally forced to layer my spam filters. Blah. The problem with a baysian spam filter (ala bogofilter, which I use) is that it can be fooled by this:
her<!5916>e to remo<!20491>ve
The number inside the angle-brackets with the bang on the front is an html comment. Since bogofilter doesn’t remove html comments, the “Click here to remove your email address” line doesn’t get parsed that way. Fortunately, this same mail was classified as spam with a quite high number (7.7) by spamassassin. So, now I’m running through bogofilter and spamassassin (with threashold of 7.5), and if either of them says it’s spam, it’s spam. Hopefully, this will cut down on the spam leaking through, without causing more non-spam to be caught. I also added a custom rule to spamassassin to give Linux Kernel Mailing List mail a -5. I had the problem before that spamassassin was classifying LKML mail as spam when it wasn’t. This should solve that problem. So, I guess, after some tuning of my threashold, I should be mostly spam free. Yay!
Discussion Area - Leave a Comment