You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Jason Haar <Ja...@trimble.co.nz> on 2005/05/30 00:22:53 UTC

Anyone know what Microsoft's "Intelligent Message Filter" does WRT tagging?

I'm wondering if Microsoft's own anti-spam solution for Exchange "tags" 
mail via headers (they call it "spam confidence level")? I can't see any 
references to actual headers being added - they may not of course - it 
could all be internal to Exchange.

However, if they do add headers, then I wonder if they'd listen to 
external ones? i.e. SAs "X-Spam-Level:",etc headers.

Does anyone have any real exposure to IMF especially with integrating a 
current SpamAssassin frontend into it?

-- 
Cheers

Jason Haar
Information Security Manager, Trimble Navigation Ltd.
Phone: +64 3 9635 377 Fax: +64 3 9635 417
PGP Fingerprint: 7A2E 0407 C9A6 CAF6 2B9F 8422 C063 5EBB FE1D 66D1


Re: Anyone know what Microsoft's "Intelligent Message Filter" does WRT tagging?

Posted by David Brodbeck <gu...@gull.us>.
Matt Kettler wrote:
> I highly doubt a MS product would take advantage of results from another
> product.

On the other hand, IF they're using statistical scoring, and IF they
include the headers in the score, then you might be able to just tag
suspected spam with a header.  Eventually the system would learn that
messages with that header had a high spam probability.  I've seen it
work with other bayes filters before.

Re: Anyone know what Microsoft's "Intelligent Message Filter" does WRT tagging?

Posted by Matt Kettler <mk...@comcast.net>.
At 06:22 PM 5/29/2005, Jason Haar wrote:
>I'm wondering if Microsoft's own anti-spam solution for Exchange "tags" 
>mail via headers (they call it "spam confidence level")? I can't see any 
>references to actual headers being added - they may not of course - it 
>could all be internal to Exchange.
>
>

I don't have any exposure to IMF, however I did a little searching..

According to this page:
http://www.microsoft.com/downloads/details.aspx?FamilyId=C1B08F7B-8CAF-4147-B074-8C9C8F277071&displaylang=en

IMF is based on "SmartScreen" technology:

http://www.microsoft.com/presspass/features/2003/nov03/11-17spamfilter.asp

Which is a statistical probability based filter that uses learning. It 
generates a probability based score.

All in all it sounds like it's very similar to bayes filter, although I 
suspect they're using a non-bayes combining algorithm. Not that this is 
uncommon, most "bayes" filters, including spamassassin bayes subsystem, 
aren't strictly bayes based.

What exact algorithms they use, and how much if any of it is a direct copy 
of methods by Paul Graham or Gary Robinson I don't know.

>However, if they do add headers, then I wonder if they'd listen to 
>external ones? i.e. SAs "X-Spam-Level:",etc headers.


I highly doubt a MS product would take advantage of results from another 
product. That's a very un-Microsoft thing to do. Usualy if MS produces a 
product in a market, they want you to use their solution exclusively.

>Does anyone have any real exposure to IMF especially with integrating a 
>current SpamAssassin frontend into it?

No, and although it may be true, I'm skeptical Microsoft allows you such power.