ietf-asrg
[Top] [All Lists]

RE: [Asrg] 2.a.1 Analysis of Actual Spam Data - Titan Key reduces spam attacks

2003-08-01 02:44:59
MessageI didn't intend "ropy" to mean anything more than "layman" or
"amateur", I'm sorry if you took as meaning something I didn't intend.

I'm afraid I can't honestly call myself a statistician as it's getting on
for 40 years since I did any academic statistics and although I have used it
sporadically since I'm not really competent any more.

Comparing against addresses with similar prior exposure using different
anti-spam techniques would be very useful.  As there's no clear upward trend
in the data, I suspect Titan Key is actually doing something useful.
However, behavioue for addresses with no current exposure is only one
measure of an anti-spam system; nehaviour for addresses that have ongoing
current exposure (displayed on websites, listed in directories, used for
posting to bitnet mail lists and usenet news groups) is also important and
may be very different.

I think if you graph your data using 7 day and 14 day moving averages you
will be able to see more clearly what is happening - the avegaing eliminates
a lot of the "noise" of day to day fluctuations.

Looking at your data, the way I would describe it is that it started out at
bit under 40, rose fairly quickly over a couple of weeks to about 50 where
it remained fairly steady for a few weeks, then fell over a couple of weeks
to about 35, where it was steady for a short time but soon began a slow rise
to a bit over 40 and has been pretty stable at that level since. It may be
about to turn down again, or turn up again, or stay fairly stable where it
is - but where it is now is a little higher than where it started, not a
little lower. On that kind of performance I wouldn't want to predict a long
term trend - the envelope may be periodic with a period slightly longer than
we have data for (a fairly long term average has gone up and down and up
again , which may be noise or may be an indication of a cycle) so fitting a
straight line to the data is not productive.

I don't have other unexposed addresses to compare your data with; highly
exposed adresses without something like Titan Key have (in my experience at
least) strong upward spam trends, but there's no knowing on what data I have
whether that's an effect of Titan key or an effect of the different
exposures.

Anyway, keep up the good work - all data is useful if we want to understand
the effects of various anti-spam measures.

Tom


  -----Original Message-----
  From: Peter Kay [mailto:peter(_at_)titankey(_dot_)com]
  Sent: 31 July 2003 21:07
  To: Tom Thomson; asrg(_at_)ietf(_dot_)org
  Subject: RE: [Asrg] 2.a.1 Analysis of Actual Spam Data - Titan Key reduces
spam attacks



  if you discard days 94 and 106 there is still a downward trend. Now, let
me be the first to raise my hand to say I'm not a statistician. So perhaps
your application of "ropy" might be more accurately expressed as "layman".
  But the numbers are there for everyone to see. And please, if there's
someone here who can honestly call themselves a statistician please step
forward and lend us your wisdom.

  The trend is either up, flat, or down. According to the linear regression
trendline, the trend is down. If there is a statistical definition of
"significant" then of course lets apply it. But if we're going to apply it
then lets also apply this trend against ANYONE elses trend that is not using
Titan Key.

  And then decide if one using a Titan Key email address reduces their spam
over time vs. someone that uses any other anti-spam approach.

  Peter





    -----Original Message-----
    From: Tom Thomson [mailto:tthomson(_at_)neosinteractive(_dot_)com]
    Sent: Thursday, July 31, 2003 9:19 AM
    To: Peter Kay; asrg(_at_)ietf(_dot_)org
    Subject: RE: [Asrg] 2.a.1 Analysis of Actual Spam Data - Titan Key
reduces spam attacks


    eems to me that you have done some pretty ropy analysis.

    did you discard the obviously anomalous outlier at day 106?

    I did some quick analysis after discarding day 94 and day 106.
Conclusion: no significant trend.  There may be a slight tren for the counts
to increase, but it certainly isn't something I would bet on from this data.
I certainly wouldn't claim a decreasing trend.

    Tom


      -----Original Message-----
      From: asrg-admin(_at_)ietf(_dot_)org 
[mailto:asrg-admin(_at_)ietf(_dot_)org]On Behalf Of
Peter Kay
      Sent: 31 July 2003 19:32
      To: asrg(_at_)ietf(_dot_)org
      Subject: [Asrg] 2.a.1 Analysis of Actual Spam Data - Titan Key reduces
spam attacks


      I've posted an updated log file analysis of spam sent to
peterk(_at_)cyber-hawaii(_dot_)com which is an old email address of mine that 
is still
in use but under the protection of The Titan Key.

      that log file is at: http://www.titankey.com/misc/logs/logPK.xls

      We're happy to report that with about 100 days of analysis, the trend
of spam attacks to this email address is DOWN.  To our knowledge this is the
only anti-spam software that has demonstrated this capability.

      As we've mentioned before, our technology responds with a 550 no such
user after analyzing the MAIL FROM and RCPT TO against a consent database.
This has the dual benefit of saving bandwidth (email contents are never
sent) and, as shown above, reducing the attacks over time.

      Peter