[Declude.JunkMail] Positive attributes (was Test suggestion request for comments...)

2003-12-08 Thread Keith Purtell
Kami:

I've been working along similar lines but run into some difficulties. I did succeed in 
getting a
group of managers and team leaders to suggest a list of good words which are often 
used in
correspondence, and are now given negative weight. However, there is an occasional 
nightmare
scenario that goes like this...

Salesperson Jones meets a potential client. Sales meeting goes well. Client seriously 
considering
purchase of new system costing more than $10,000. Client sends email from a Yahoo 
account to
Salesperson Jones containing text like this: Very interested in what you described. 
Please call me
at 10 a.m. to discuss purchase. This incoming mail doesn't contain any of our good 
words, and it
comes from a common spam source, has other possible flaws, etc. So it gets tagged as 
spam and
delivery is delayed.

The only quick fix we've found is to assign a large negative weight to email addressed 
to a few
in-house accounts. Of course it's not really a fix if they now get a ton of spam. I've 
suggested a
special word that could be distributed to potential new clients for use in sending 
us email, but
the idea has not caught on.

Keith Purtell, Web/Network Administrator
VantageMed Operations (Kansas City)

CONFIDENTIALITY NOTICE: This email message, including any attachments, is for the sole 
use of the
intended recipient(s) and may contain confidential and privileged information. Any 
unauthorized
review, use, disclosure or distribution is prohibited. If you are not the intended 
recipient, please
contact the sender by reply email and destroy all copies of the original message.


 -Original Message-
 From: [EMAIL PROTECTED]
 [mailto:[EMAIL PROTECTED] Behalf Of Kami Razvan
 Sent: Sunday, December 07, 2003 12:55 PM
 To: [EMAIL PROTECTED]
 Subject: RE: [Declude.JunkMail] Test suggestion  request for
 comments...


 Bill..

 This goes well along the line of the subject that was
 discussed a while back
 and one that could help a great deal.

 Right now we are concentrating on negative aspects of the email - to
 minimize FP and even further reduce CPU we should give some
 attention to
 some positive aspects as well.

 If we can identify the positive attributes correctly we can
 further tighten
 our filters and be more generous with weighing the negative
 attributes.

 A discussion on positive traits could be a good start  I
 second the motion.

 Regards,
 Kami

 -Original Message-
 From: [EMAIL PROTECTED]
 [mailto:[EMAIL PROTECTED] On Behalf Of Bill Landry
 Sent: Sunday, December 07, 2003 1:32 PM
 To: [EMAIL PROTECTED]
 Subject: [Declude.JunkMail] Test suggestion  request for comments...

 Scott, you have probably seen requests like this before,
 however, I think
 this would be a great way to support most corporate and some
 ISP e-mail
 domains with a negative weight based test:

 HELO  RDNS domain match -5
 HELO  RDNS  MAILFROM domain match -10
 HELO  RDNS domain match  IPINMX -10 (yes, IP-in-MX) HELO 
 RDNS  MAILFROM
 domain match  IPINMX -15 or ENDALLTESTS

 I say domain meaning just the last two segments of the
 FQHN, that portion
 that is registered with domain registrar.  Since all of these
 tests are
 already run by Declude, if a bit of logic could be added to
 support a test
 like this, I think it could help us get a lot of legitimate
 mail delivered
 with fewer held due to FPs.

 Also, if people feel that the last test above is a very good
 indicator of
 legitimate e-mail, then if this test is run first (before all
 other tests),
 and there is a match with the last test shown above, and
 there was variable
 to ENDALLTESTS (and deliver), then this would also cut down
 on processing
 requirements.

 Thoughts anyone...?

 Bill


---
[This E-mail scanned for viruses by Declude Virus]

---
[This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type unsubscribe Declude.JunkMail.  The archives can be found
at http://www.mail-archive.com.


RE: [Declude.JunkMail] Positive attributes (was Test suggestion request for comments...)

2003-12-08 Thread Kami Razvan
group of managers and team leaders to suggest a list of good words which
are often used in correspondence,

Hi Keith:

Actually I think what Bill was originally talking about and what I was
trying to say was a way to actually credit good servers.

Lets say..

- No blacklists
- Valid REVDNS
- Valid Helo
- Valid looking email

 other attributes that are purely server related.

If all valid then one can assign a negative weight to.  This would be to
credit good servers.

Yahoo, Hotmail, etc. are always going to be problematic and I am not much
concerned about them.  My concern is if we can assign credit to good ones
then our other filters can be a little more loosely defined.

Regards,
Kami


---
[This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type unsubscribe Declude.JunkMail.  The archives can be found
at http://www.mail-archive.com.


RE: [Declude.JunkMail] Positive attributes (was Test suggestion request for comments...) request for comments...)

2003-12-08 Thread Colbeck, Andrew
And to state the obvious...

Giving negative weights to combinations of good server configurations
would benefit correspondence from legitimate mail servers that would
otherwise get tagged as spam by false positives on content filters, e.g. a
joke e-mail with bad words or a newsletter with URLs that contain
unnecessary IP addresses or escaped characters.

Kami suggested multiple different combinations that could be termed good,
which means that a legitimate mail server wouldn't need to be perfect in
every way in order to be scored good.

My own suggestion is combine the client source and the mail server with some
logic.  For example, if a message comes to my mail server from a mail server
at c-67-160-69-182.client.comcast.net it is very likely to be spam; however
if the source is c-67-160-69-182.client.comcast.net and the mail server is
sccrmhc13.comcast.net then it likely to be legitimate, even though both
sources are likely to be listed in multiple ip4r databases*.

* Why?  An open proxy or whatever on a consumer broadband network is
unlikely to be a valid smtp server, and the whole range is in many
blacklists.  However, a message that comes from that range through the smtp
server at that ISP is likely to be normal consumer mail.

Going a little further, a traveller on a laptop as a guest at some random
ISP but using his smtp server at his own ISP is also going to look bad, but
is not guaranteed spammy.  If that guest IP is also in spamcop or some
other fluid ip4r database, the balance of tests should push the message over
the weight limit.

Andrew 8)

-Original Message-
From: Kami Razvan [mailto:[EMAIL PROTECTED] 
Sent: Monday, December 08, 2003 2:11 PM
To: [EMAIL PROTECTED]
Subject: RE: [Declude.JunkMail] Positive attributes (was Test suggestion 
request for comments...)


group of managers and team leaders to suggest a list of good words which
are often used in correspondence,

Hi Keith:

Actually I think what Bill was originally talking about and what I was
trying to say was a way to actually credit good servers.

Lets say..

- No blacklists
- Valid REVDNS
- Valid Helo
- Valid looking email

 other attributes that are purely server related.

If all valid then one can assign a negative weight to.  This would be to
credit good servers.

Yahoo, Hotmail, etc. are always going to be problematic and I am not much
concerned about them.  My concern is if we can assign credit to good ones
then our other filters can be a little more loosely defined.

Regards,
Kami


---
[This E-mail was scanned for viruses by Declude Virus
(http://www.declude.com)]

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type unsubscribe Declude.JunkMail.  The archives can be found
at http://www.mail-archive.com.
---
[This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]

---
This E-mail came from the Declude.JunkMail mailing list.  To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type unsubscribe Declude.JunkMail.  The archives can be found
at http://www.mail-archive.com.