I've been seeing more and more spam using sequences of garbage character strings; I'm guessing to get around Bayesian filtering? Also the obfuscating comments seem to contain garbage strings, but the existing rule finds those pretty well.
Is anybody working on a rule for lots of garbage words in email? Now, if I were a spammer I'd pick real words out of a dictionary, which is easy enough to do, but since they *do* seem to be using strings of garbage words at the moment.... --
Garbage character strings are mostly for avoiding message-hash type systems like razor, dcc and pyzor. They have little effect on bayes.
There are a lot of spams out there that do use real dictionary words as an attempt to evade bayes.. they mostly use words which are large and not commonly used in non-academic settings. They tend to burry them hidden in html comments or hidden frames.
------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk