Incidentally, I've noticed a worrying difference between email spam and weblog spam. With email spam, the spammer can maybe approximate your spam list (after all, even spammers get spam), but they can't see your non-spam list without cracking your email account. On the other hand, with weblog spam no one sees the spam you deleted, but all the non-spam comments are right there for everyone to see and parse. It would be trivial for a comment spammer to post an exact repeat of someone else's comment but with Viagra links all over the place. In the long run, a Bayesean filter for comment spam might degenerate into an IP and URL blacklist. Is there any algorithmic defense against an attacker who has access to everything previously blessed by the algorithm?

![5 comments [Comments]](/nb/assets/comments.png)
Posted by Jeremy Penner at Tue May 18 2004 02:28
Posted by Leonard at Tue May 18 2004 03:14
Posted by josh at Wed May 19 2004 21:20
Posted by Leonard at Thu May 20 2004 01:15
Posted by josh at Thu May 20 2004 11:44