From: "John Andersen" <[EMAIL PROTECTED]>
On Saturday 15 July 2006 03:08 am, Loren Wilton wrote:
and if spammers weren't so
incompetent in general it would be even harder than it is.
An odd comment, especially for a project like Spamassassin which
has had to run full out for the last dozen years just to remain
in a dismal second place compared to spammers.
Dismal second place?
First, where are you coming from? Are you a large ISP using global
everything and have to be lenient and rely on signatures and BLs for
most of the good stuff? Even that gives on the other of a 10:1 ratio
for killed spam to escaped spam.
And what do you figure would be better than dismal second place given
that the miss to kill ratio even on a poorly configured out of the box
SA runs better than 1:1 and probably better than 3:1? SA is forcing the
spammers to be more creative and mangle their message ever more to get
it across. The mangling is counter productive. It telegraphs "This is
spam, delete it" to the recipient even without decoding the words. The
image only trick is a relatively new one and modestly effective. But
all image/no message in itself is good spamsign. Random words in the
message are Bayes food. They make Bayes work better.
On the whole I'd say I am winning. I have 7 escaped spams out of roughly
8500 spams total. Two of those were plain empty. The rest were new
formats. (An electronics parts based 419, for example.) They are all
unique in one way or another. Bayes actually adapted to catching them.
Somehow I figure a better than 1200:1 scoring ratio is a pretty lopsided
win for SpamAssassin. (I have NO other spam filtering involved, not even
grey listing, which is the other really good tool.)
Are you trolling?
{^_^}
