OT: spam filtering idea

Anthony Baxter anthony at interlink.com.au
Wed Jan 15 04:27:12 CET 2003


>>> Tim Peters wrote
> [Paul Rubin]
> > ...
> > Spambayes is already working better than spamassassin?  Wow.
> 
> It depends on what you use it for.  It was intended to be used by a single
> person on their own email, and it quickly learns so much about a single
> person's quirks that even very early versions of the spambayes code did at
> least as well as a well-maintained SpamAssassin.

I know that for _my_ use, it kicks spamassassin's butt.

I still have spamassassin running over the inbound email (but in non-mangle-
the-spam mode), so I can check occasionally how it's going. 

For the last 100 spams in my spam folder, spambayes nailed 92 as spam,
and 8 as unsure. There were no false negatives (missed spam), although
I seem to see about 1 a week or so of those, and I've yet to get a 
false positive. SA tells me that 86 of these 100 are spam, and 14 are not.
When I was using SA regularly, I had to put an enormous set of whitelist
addresses to let things through like Blackstar's regular mailouts and
other commercial email that I wanted to see, as well as for things like
RISKS digest (which, for some reason, SA _really_ hated). This is 
obviously a hole that spammers will try to exploit - I already see 
spams with sender addresses @amazon.com, presumably to try and slip 
through this sort of hole.

Hopefully SA will pick up some of the techniques in a future version.

Anthony






More information about the Python-list mailing list