[Spambayes] Spam which made it through the filter

Meyer, Tony T.A.Meyer at massey.ac.nz
Tue Sep 2 20:05:55 EDT 2003

> I believe that SpamBayes has to eventually dig into Received 
> headers, and check the address where the spam originally came 
> from (convert the ip address to host name perhaps).

The Received headers can be used now, although there isn't any lookup.
(I'm still getting around to the results from the other lookup...)

> The message below has (in addition to $RANDOMIZE bug that 
> spammers will fix) many words that I never get in any of my
> good e-mails. But the problem is that if the spammer has a
> huge dictionary, which is likely, then learning these does
> not do much good.

I think a message_length token would be more effective against this sort
of mail.  Or even something funky like the ratio of hapaxes to
non-hapaxes (or to message length).  I don't get enough of these (in
fact, I haven't got any fn's yet) to be able to test this, though.

=Tony Meyer

More information about the Spambayes mailing list