using sa-learn on spam attachements

[Ed Bruce]

    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "US-ASCII" character set.  ]
    [ Some characters may be displayed incorrectly. ]

Greg Matthews wrote:

 On Thu, 2005-09-15 at 08:25 -0400, Ed Bruce wrote:
  

 I use MailWatch.This has a nice web based interface to submit email to
the SA-learn function.
    

 I use mailwatch too but I dont have the time or resources to spend on
checking spam/notspam accuracy and feeding the inaccuracies back to the
bayes...

I can ask my users to "forward as attachment" the mails that get marked
wrong but even if I could trust them to do this right I'd still have to
process the mails before "learning" from them. (admittedly, if I could
guarantee that every user followed the instructions correctly, some of
the work could be automated, but...)

The only learning material that I can trust 100% is my own FPs and FNs
but then I have to strip out the attachments on the FPs and strip out
the MS headers from the FNs before I can use them (as I understand it).

GREG
  

This is what I currently do, I have users send email to one of two
special mail drops (one for spam and the other for ham) and I also do
what someone else suggested I get an archive copy for 7 days. I then use
the forwarded email with the Reports web page in MW to find the email and
feed it to SA-learn. This way I don't have to strip off the headers.

------------------------ MailScanner list ------------------------
To unsubscribe, email jiscmail at jiscmail.ac.uk with the words:
'leave mailscanner' in the body of the email.
Before posting, read the Wiki (http://wiki.mailscanner.info/)
and the archives (http://www.jiscmail.ac.uk/lists/mailscanner.html).

Support MailScanner development - buy the book off the website!