Hello everyone,

I received a very obvious spam in my Inbox today, and I was surprised
that SpamAssassin's Bayes didn't catch it.  Examining the headers
revealed that indeed Bayes was not consulted:

X-Spam-Status: No, hits=1.3 required=1.5 tests=HTML_30_40,
HTML_FONT_INVISIBLE, HTML_MESSAGE autolearn=no version=2.63

I noticed that the spam was addressed to two people at my domain - not
just me - and I'm using per-user Bayes via spamc.  So I edited the
message to take out the other recipient and re-injected the message into
the queue.  Bayes gave it a confidence of 99; the message ended up with
6 points total.  

Can qmail-scanner be configured to expand recipient addresses and call
spamc for each recipient?  Or will this require patching?  If it will
require patching... is it a good idea?  Or would it be vulnerable to
some kind of DoS (ie. send a mail with 500 recips only costs ~200kb of
bandwidth, but internally qmail-scanner will expand it into 100,000kb
for qmail to deal with).  

-- 
[EMAIL PROTECTED]

Administrator, tgpsolutions
http://www.tgpsolutions.com

Attachment: signature.asc
Description: This is a digitally signed message part

Reply via email to