max at inmindlabs.com
Wed Nov 14 19:58:23 GMT 2012
I set up a new MailScanner system with the latest software, version of
SpamAssassin with Bayes, etc. a couple of month ago. It was working
really well for a while, but in the last week we've started getting
bombed with loan, medicine spam, etc. The most prevalent spam email of
late is the one that has wide but short image that tries to present
itself at the top of the email. All the images look the same size across
different spam topics and different senders, but of course Outlook
blocks the image with a red x.
Well when I look at the score, most have a fairly high score except for
the Bayes which will usually be at 50%. So this causes it to be
considered non spam. I send tons of these for learning, but no change.
The next one will come in with a 50% bayes score. At this point is this
considered Bayes Poisoning? Should I start over on Bayes? I manage the
system, so I also have an IMAP account to the MS system and created a
program that grabs the email out of my IMAP spam folder, sends it for
learning, and then saves it in an mbox file. So I have every spam
message I've ever sent. So would the best course at this point be to
start clean on Bayes and then feed all the spam back in?
FYI, this MS system does filter tons of spam. Yesterday's count was
6,702 spam caught. But I've seen an increase from 1 or 2 a day
personally to about 10 - 20, some of them going to Outlook Junk.
As I'm writing this I just got another one of those image spams about
medicate. Here is the score. If my Bayes was reporting right, it would
have been considered spam.
not spam, SpamAssassin (not cached,
score=4.733, required 5, BAYES_50 0.80, HTML_MESSAGE
RDNS_DYNAMIC 0.98, URIBL_DBL_SPAM 1.70, URIBL_JP_SURBL
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the MailScanner