Hello Sniffer folks,

  Just in case you've been watching, the last few days have seen a
  large number of new rules... However, this time, it's not a spam
  storm.

  (From Rule Base Change Rate Analysis>

  <http://www.sortmonster.com/MessageSniffer/Performance/ChangeRates.jsp>
  
  Days Ago Adjustments
  -------- -----------

  0        1322
  1        2509
  2        381
  3        494
  
  We have been experimenting with some new robots with improved
  intelligence. These robots help us to generate new rules
  automatically so that we can respond very quickly to new spam. They
  are getting very smart, but we always review the rules they create
  manually in order to maintain their quality.

  During the last 15 hours or so, we've been running a "deep scan"
  experiment that allows us to capture additional rules that might
  exist in spam we've already filtered out. While tuning the robots to
  perform deep scans we've allowed them to run "wide open" and in the
  process we discovered that they can be quite prolific.

  While we continue our tuning process we're going to see some unusual
  numbers in the Rule Base Change Rate analysis. We should be done
  with this work in the next few days. After that things will settle
  down and look more "steady".

  The new robots have already had a very positive effect on our
  capture rates. The combination of changes we've implemented over the
  past six weeks has significantly increased Message Sniffer's
  accuracy in all rule groups - driving false positives to record lows
  and capture rates to record highs. We expect to see these trends
  continue over time as our robots and SPHUD tools become smarter.

  (SPHUD = SPam Heads Up Display: This is a tool we use to research
  new spam that arrives at our system through spamtraps and user
  submissions.)

  A good place to see the effects of our work is on the Spam Test
  Quality Analysis page by Markus Gufler:

  <http://www2.spamchk.com/public.html>

  I hope you find this information to be both useful and interesting.

Thanks,
_M

Pete McNeil (Madscientist)
President, MicroNeil Research Corporation
Chief SortMonster (www.sortmonster.com)



This E-Mail came from the Message Sniffer mailing list. For information and 
(un)subscription instructions go to 
http://www.sortmonster.com/MessageSniffer/Help/Help.html

Reply via email to