towards a better comments spam filter
Oh and post comments here to make sure I haven't broken it for everyone :-)
--Kirk Fri, 15 Jul 2005 08:00:51 -0400
Oh, and one silver lining to this: having to go over old "real" comments to weed out the spam made me realize that I really do appreciate how much the comments feature has addded to this place...
--Kirk Fri, 15 Jul 2005 09:30:08 -0400
Ever consider Bayesian filtering for comments?
--Jeremy Penner Fri, 15 Jul 2005 10:40:56 -0400
Well, that's one of the downsides of homebrew software, either having to roll your own or integrate w/ someone else.
Comments spam seems much less sophisticated than its email brethren, so at least for now that kind of filter might work really well. Still, I'll wait 'til my simpler methods fail before advancing the arms race like that.
--Kirk Fri, 15 Jul 2005 10:47:11 -0400
Wow! I totally didn't understand the jargon side of things even though I was shocked by just how technical and wide ranging SPAM has gotten. Bastards!
--Mr. Lex Fri, 15 Jul 2005 11:47:44 -0400
Here is a thought on the multiple-textareas idea, if 2 are hidden by CSS, then if there is content in either of them the comment is spam. Since the script won't know which ones is the real ones you'd be good. I doubt the script would be smart enough to put content in only one of them, but if so, then there is only a 1 in 3 chance of it showing up. So then you add 10 hidden textareas.
--Mr. Ibis Fri, 15 Jul 2005 13:33:39 -0400
Yeah, I guess CSS is
A. pretty standard now, at least for hiding things
and
B. more non-trivial to parse than, say, javascript, so it might be a while before they can put in the condition "ignore form elements that are hidden or in containers hidden by CSS"
--Kirk Fri, 15 Jul 2005 13:36:17 -0400
thanks for the shout out, yo.
--FoSO Fri, 15 Jul 2005 17:22:36 -0400

Comments Disabled... (Thanks Dirty Rotten Spammers)
Feel free to write kirkjerk at gmail dot com!