Charles Leeds wrote:
It seems like doing a parsing of two letter tokens for probability in a
From: address would be very effective.
[EMAIL PROTECTED]
yf
fq
qs
sm
mt
tf
fx
xx
That right there should identify the message as spam without ever
looking at the message.
As I understand dspam, a letter combination doesn't constitute a token
(apart from any number of letters being part of any single such).
As we have dspam running, its accuracy "as is" is 99+% over around
90,000 inbound messages (dspam_stats -H).
What's your point? Isn't dspam as is accurate enough for you?
--Tonni
--
Tony Earnshaw
Email: tonni at hetnet dot nl
