On Wednesday, March 30, 2005, 10:35:52 PM, Darin wrote:

DC> Pete,
DC> �
DC> Have�you make�significant changes to the  sniffer rulebase in the past 
couple of days?
DC> �
DC> I'm seeing a _huge_ reduction in hold queue  messages...
DC> roughly down 65%...�while total message volume is steady.�  Only
DC> thing I can figure is that the rulebase is suddenly identifying
DC> most of the  messages that fail other tests as well, pushing most
DC> over the delete limit....  or other tests like SpamCop,
DC> Mailpolice, etc.�have made significant  changes...� I've checked a
DC> few sites for news, but am not seeing anything  new.
DC> �
DC> The sudden change has me a wee bit  concerned...cautiously optimistic, but 
concerned.

THis might be better asked on the Sniffer forum rather than Declude's
though I'm sure they don't mind.

The only thing I can think of is that there has been a greater use of
message fragment rules over the past few days in response to some of
the newer campaigns. I wouldn't call that a radical change - but it
has been a moderately heavy shift. In particular there is a new snake
oil campaign that is using a number of randomized obfuscated segments
in their message and we've been capitalizing on that.

I don't see any significant shifts in the statistics. What I do see is
a subtle change in the shape of the new rule capture curve (see the
left side of this chart):

http://www.sortmonster.com/MessageSniffer/Performance/ChangeRates.jsp

I have also seen higher spam rates and SNF capture rates in recent
MDLP data on our system:

http://www.sortmonster.com/MDLP/MDLP-Example-Short.html

What counts in cases like these are false positive rates... If it
seems that we're catching a lot more spam then lets be sure it really
is spam. So far FP rates are nominal though there is a spike yesterday
in the number (this appears to be an automated system that submits FPs
from users -- the batch contains a larger than usual number of
duplicate submissions -- this happens from time to time with this
customer).

http://www.sortmonster.com/MessageSniffer/Performance/FalseReportsRates.jsp

Hope this helps,

_M

  


---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type "unsubscribe Declude.JunkMail".  The archives can be found
at http://www.mail-archive.com.

Reply via email to