>if may be an idea to change
>the code (not now, but... in a while) so that HMM will perform a first
>check,
This is the plan. If HMM gets no result or even one based on a single match (or
two or three) - Bayes should do the detection.
Now both engines are running, if configured and each is scoring. Possibly it is
also a solution to set the valence values in a way, that a mail is blocked if
both engines detecting spam.
Thomas
-----"GrayHat" <gray...@gmx.net> schrieb: -----An:
<assp-test@lists.sourceforge.net>
Von: "GrayHat" <gray...@gmx.net>
Datum: 25.10.2011 14:41
Kopie: "Thomas Eckardt" <thomas.ecka...@thockar.com>
Betreff: Re: Antwort: Re: [Assp-test] fixes, changes and news in assp 2.1.2
build 11295
> What I currently need are some reports about the detection behavior of
> the HMM. I saw several mails, which where different detected by Bayes
> and HMM : Bayses: spam - HMM ham or via vers. In all those cases
(100%) HMM
> was right and Bayes was wrong - but this should be normal, because HMM
> is much more complex -> much more exact. Currently the HMM uses a very
> flat logic - to hold the memory usage and runtime as low as possible.
> The number of tables required depends on the logic - so we first need
> to find the best logic, before we can improve the rest.
I may try running HMM on some boxes and possibly feed you some
data (feel free to send me more infos about the kind of data/infos you
may need); in such a case, I'd start by setting HMM to "monitor" so that
it will be able to gather enough data; wait for the next "rebuildspamdb"
and then switch it to "score"; may that be ok for you ? Also, and about
HMM and Bayes; I think that the two may be complimentary; I mean..
given that HMM is overall more accurate; if may be an idea to change
the code (not now, but... in a while) so that HMM will perform a first
check,
then, in case of doubt or if the message is below a given size or in any
case doesn't have some "characteristics", the message may be fed to
the bayes filter for a "second opinion"; this means that, on the long
run,
most messages will be categorized by the HMM while some of them
will still step down to the bayes for some additional checking; a
possible
example may be the messages flagged by other previous filters but not
rejected :) - the basic idea is to avoid running both HMM and Bayes if
possible and using bayes only as a "second check" in case of doubt
DISCLAIMER:
*******************************************************
This email and any files transmitted with it may be confidential, legally
privileged and protected in law and are intended solely for the use of the
individual to whom it is addressed.
This email was multiple times scanned for viruses. There should be no known
virus in this email!
*******************************************************
------------------------------------------------------------------------------
The demand for IT networking professionals continues to grow, and the
demand for specialized networking skills is growing even more rapidly.
Take a complimentary Learning@Cisco Self-Assessment and learn
about Cisco certifications, training, and career opportunities.
http://p.sf.net/sfu/cisco-dev2dev
_______________________________________________
Assp-test mailing list
Assp-test@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/assp-test
