Hi folks,<snip>
I just received the attached maessage, it tripped SA at 5.6 but the only thing that got hit was the BL rules. How did this get such a low score?
version=2.53
First, a side-note.. are you aware of the dreaded bug 1589, that allows spammers to construct messages with huge negative scores? This bug affects 2.50-2.53.. you might want to consider an upgrade if you have problems, unless you've hack fixed it yourself by zeroing out the affected rules.
As for how it got a low score.. simple.. string obfuscation with @'s to avoid matching SA's string searches. The bayes engine can be trained really well on this kind of stuff as it's really uncommon if not nonexistant for normal people to use words like "[EMAIL PROTECTED]".
At the time SA 2.53 came out, w4r3z speak type spam was still new, and there were no rules to catch them. I think there are a few that cover some of this stuff in 2.60, but I don't think it's very comprehensive. Bayes can still fill in a lot of gaps here, and is really a very powerful way to deal with adapting trends in spam.
------------------------------------------------------- This SF.Net email sponsored by: Free pre-built ASP.NET sites including Data Reports, E-commerce, Portals, and Forums are available now. Download today and enter to win an XBOX or Visual Studio .NET. http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
