Filter breakers

[David Relson]

Hi Stephen,

What you've got sounds very thorough!  Since what you're doing with
"-n" and "-s" is just what "-u" is designed for, an optimization
(simplification) of your process is possible.

It also seems like bogofilter is run after the others so few messages
get actually go through the "-n" / "-s" registration process, which
would explain the low token count for head:Apr.

Enjoy!

David

On Sat, 5 Apr 2008 11:07:08 +0930
Stephen Davies wrote:

> I use Amavis, Amavisd, Clamav, Bogofilter and milter to scan all mail.
> 
> Based on the bogofilter return value, I run bogofilter -n or
> bogofilter -s to register the result and if it is spam, I put it in a
> mailbox called spambox for review.
> 
> I haven't had a false "spam" in ages; just incorrect "hams".
> 
> When a spam gets through to my inbox, I forward it to user spam which
> is an alias for a shell script that runs bogofilter -Ns (twice) to
> reverse the registrations.
> 
> I am going to follow your advice regarding registering a bunch of
> stuff to see if it improves things.
> 
> Cheers and thanks,
> Stephen
> 
> On Friday 04 April 2008 21:03, David Relson wrote:
> > On Fri, 4 Apr 2008 18:10:13 +0930
> >
> > Stephen Davies wrote:
> > > G'day David.
> > >
> > > I hadn't checked the month total frequencies. I get:
> > >
> > > X-Bogosity: Ham, tests=bogofilter, spamicity=0.000006,
> > > version=1.1.5
> >
> > ...[snip]...
> >
> > > When did headers start being included? Probably the bulk of my
> > > database is several years old. Do you think something like
> > > bogoutil -m wordlist.db -a 20050101 or bogoutil -m wordlist.db -c
> > > 500 might help?
> > >
> > > Cheers and thanks,
> > > Stephen
> >
> > Header tagging was implemented about 5 years ago.  Do you, per
> > chance, have it disabled via the "-H" flag?
> >
> > Personally, I use tristate classification and database autoupdating
> > ( "-u" flag).  This involves additional care and overhead because
> > it's important to correct classification errors (because one error
> > can lead to additional errors if not corrected).  I also filter
> > messages classified "Unsure" to a special folder so they can be
> > registered appropriately (as ham or spam).
> >
> > Another idea is to take a day's incoming messages and register them
> > all.  As the "head:Apr" token will be in all the ham and spam,
> > this will neutralize the token.  Repeating this exercise once
> > per month will neutralize all the month tokens.
> >
> > Regards,
> >
> > David
> 
> -- 
> ========================================================================
> This email is for the person(s) identified above, and is confidential
> to the sender and the person(s).  No one else is authorised to use or
> disseminate this email or its contents.
> 
> Stephen Davies Consulting                            Voice: 08-8177
> 1595 Adelaide, South Australia.                             Fax:
> 08-8177 0133 Computing & Network solutions.
> Mobile:0403 0405 83


-- 
David Relson                   Osage Software Systems, Inc.
relson at osagesoftware.com       Ann Arbor, MI 48103
www.osagesoftware.com