Re: Mark an email as junk because of the TYPE of email, not the domai.

From: Vanguard (use_ReplyTo_at_domain.invalid)
Date: 02/18/05 

Date: Thu, 17 Feb 2005 23:59:08 -0600

"MaxBro" <maxbro@mindspring.com> wrote in message
news:uF$3P2VFFHA.2824@tk2msftngp13.phx.gbl...
> Sorrrrrrrry, lol. I do detect a case of ATTITUDE in your response.
> The program I was referring to is I Hate Spam which I was using the
> "full" versions of both for OE and Outlook2003.
>
> Sounds like you might want to take a break from running throught the
> newsgroups.......just a suggestion.

You still never mentioned WHICH version of Outlook that you use. As
mentioned, Outlook 2003 added Bayesian filtering. From what I've heard
from other users that have tried a few Bayesian enabled products,
Microsoft's Bayesian filtering is "loose" so it isn't effective for
awhile, plus it isn't very configurable, especially to prevent
poisoning. Most Bayesian filters start out completely dumb because they
have no statistics yet on which to exercise against new e-mails. Some
let you feed in sets of good e-mails to train them which ones are not
spam and let you feed in sets of bad e-mails to train them which ones
are bad, but then you need those old e-mails to pre-train your Bayesian
filter to populate its weighted word database. Typically you need a
corpus of around 600 mails (which is a sample representing your typical
load of good and spam mails) before the Bayesian filter starts to
perform optimally. I've heard (but not confirmed) that Microsoft's
updates include a preset list of weighted words but that list is
obviously not based on your historical experience with e-mails. Is
there an option in Outlook 2003 to pre-train it based on YOUR sets of
good and bad e-mails? The idea is to provide a "starter set" in the
Bayesian database but it's not actually tailored to your experiences
with e-mails.

IHateSpam also incorporates Bayesian filtering. If you feel that
IHateSpam did a better job than the new Bayesian filter incorporated
into Outlook 2003 then why aren't you still using it? I would suspect
IHateSpam would be a better Bayesian filter than the one in OL2003.
Won't IHateSpam work with Outlook 2003? According to
http://www.sunbelt-software.com/evaluation/930/web/documents/ihs_unsupported_email_faq.pdf,
it looks to run as a proxy through which your e-mail client connects to
your mail server. If not, SpamPal makes a good alternative and it also
runs as a proxy so any POP3-compliant e-mail client can use it, and it's
free to boot.

As with all other fluff or 3rd party software incorporated into
Microsoft's products, don't expect them to perform equivalent to the
full-version products available from 3rd parties. IHateSpam, SpamPal
(free), and SpamBayes (free) are all probably better solutions than
OL2003's Bayesian filter. You might even be able to add Cloudmark's
SpamNet (now called SafetyBar) which is a user community voting scheme
to identify spam for another layer of protection. And, of course, you
should have enabled the spam filter on your ISP-provided mailbox. Don't
expect an ultimate spam solution from Microsoft. 

-- 
____________________________________________________________
Post your replies to the newsgroup.  Share with others.
E-mail reply: Remove "NIXTHIS" and add "#VS811" to Subject.
____________________________________________________________

Relevant Pages

  • Re: Mark an email as junk because of the TYPE of email, not the domai.
    ... > statistics yet on which to exercise against new e-mails. ... > feed in sets of good e-mails to train them which ones are not spam and let ... > mails) before the Bayesian filter starts to perform optimally. ... > IHateSpam also incorporates Bayesian filtering. ...
    (microsoft.public.outlook.general)
  • Re: Email Filtering
    ... That rule is NOT available in Outlook Express. ... Boy, that will delete a LOT of e-mails, most of which do NOT have a URL. ... Note that HTML-formatted e-mails can get around simplistic word lists ... Microsoft included a Bayesian filter, ...
    (microsoft.public.windowsxp.general)
  • Re: Blocked Messages
    ... a pre-defined database to its Bayes-like filter in Outlook 2003+. ... They say they support blacklists. ... identify known spam sources while your Bayesian filter catches up. ...
    (microsoft.public.windows.inetexplorer.ie6_outlookexpress)
  • Re: [SLE] sa-learn on server (Postfix/Procmail/Cyrus)
    ... > the criteria for spammyness is a company policy. ... not if you allow users to directly update the Bayesian filter ... If you want to keep the company policy, then spam should be ... and another user sending the same email to the list of ...
    (SuSE)
  • Re: Python from Wise Guys Viewpoint
    ... The difference is that I can draw one up for a Bayesian filter: ... essentially a specification how user input should be used to classify data. ... The actual behaviour of the filter for each given spam isn't specified ... unspecifiable) - but when I'm the program's maintainer, I don't care ...
    (comp.lang.lisp)