Re: consistant failures on a new kind of spam

Brian Burton <[email protected]> Wed, 15 Nov 2006 09:13:40 -0500
Newsgroups gmane.mail.spam.spamprobe.general
Message-ID <[email protected]>
Phil Hord wrote:
> Anthony Campbell wrote:
> All this means that Spamprobe is *suddenly* understanding my spam
> better.  Why?

So far for my own email I see spamprobe catching these much more often. 
  Looking at the scores it seems that -o suspicious-tags helps.  Not 
sure, but -h might be even more helpful.

Another possibility is that spammers are rotating through different 
sources of garbage text in their spams.  Some of that garbage will be 
highly hammy for some people but not for others.  Technical docs help 
against technical people's databases, but not against ordinary people's 
databases.  Also you would expect that if too many spams use the same 
set of technical terms then those terms will gradually lose their 
effectiveness over time.


> Spamprobe did not suddenly get better at detecting spams, though.
> Yesterday it still missed several dozen.  But today I've seen so few.
> Did it get better?  Or is there some holiday in Russia/China/Nigeria
> that I'm not aware of?  (nope -- spam volume is high today, around 550
> so far, well above normal for the day.)

Several dozen?  Wow!  The most I've had slip through in one day is 2 or 
three (more if duplicates of one that made it through arrive).


> Would a nearly full hash database cause Spamprobe to misidentify spams?

Nearly full?  That would certainly drop performance to the floor since 
it would create a huge number of hash collisions. Hash tables need 
plenty of empty slots to avoid collisions.  Hash tables turn into 
glorified lists with too many collisions.

If performance is poor enough procmail might time out waiting for SP 
(worst case) and deliver the email with an empty X-SpamProbe header.

If the database became completely full you would definitely have 
problems since the database could not be updated.

All the best,
++Brian

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

_______________________________________________
Spamprobe-users mailing list
Spamprobe-users-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org
https://lists.sourceforge.net/lists/listinfo/spamprobe-users
PGP.sig (application/pgp-signature, 512 B) - not displayed