You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Dave Stern - Former Rocket Scientist <da...@umiacs.umd.edu> on 2004/08/20 16:44:01 UTC

AWL on SA3.0rc1

Since installing SA3.0, I'm getting a much higher false-positive rate
that tends to come from AWL test. Sending myself a message, I get

...
X-Spam-Status: Yes, score=5.2 required=4.0 tests=ALL_TRUSTED,AWL,
     NO_REAL_NAME
    autolearn=no version=3.0.0-rc1

  0.2 NO_REAL_NAME           From: does not include a real name
-2.8 ALL_TRUSTED            Did not pass through any untrusted hosts
  7.8 AWL                    AWL: From: address is in the auto white-list

What is AWL? how can I change it?
If it's lying anyway, can I get away with rescoring this rule?
(header AWL -7.8)

TIA


  =-=-=-=-=-=-=-=-=-=-  generated by /dev/dave -=-=-=-=-=-=-=-=-=-=-=-=
  David Stern                                    University of Maryland
            Institute for Advanced Computer Studies

Re: AWL on SA3.0rc1

Posted by Theo Van Dinter <fe...@kluge.net>.
On Fri, Aug 20, 2004 at 10:44:01AM -0400, Dave Stern - Former Rocket Scientist wrote:
> What is AWL? how can I change it?
> If it's lying anyway, can I get away with rescoring this rule?
> (header AWL -7.8)

As it says in the docs, the auto-whitelist is now enabled by default
(before you needed to enable it specifically via the "-a" commandline
parameter).  See the README and/or Mail::SpamAssassin::Conf's
"use_auto_whitelist" section (there's probably something in the Wiki
about it too,) for more information.

-- 
Randomly Generated Tagline:
A short cut is the longest distance between two points.

Re: Re[2]: AWL on SA3.0rc1

Posted by sn...@fastmail.fm.
On Sun, 22 Aug 2004 10:54:14 -0700, "Robert Menschel"
<Ro...@Menschel.net> said:
> Daniel,
> 
> My primary hesitation with AWL has to do with spoofing -- we get too many
> spams "from" one of our domain addresses "to" another. I'm concerned that
> these will lead AWL to be incorrect wrt our domains. (Since our email
> runs on shared web servers, ALL email except webmail comes from external
> sources, even those I send to others within our domain.)

Bob,
If I remember correctly, the AWL database stores both the sender's
e-mail address and the IP address it was received from. So the average
of scores for messages received from "robert@menschel.net:XX.XX.XX.XX"
will be kept separate from the scores of messages received from
"robert@menschel.net:YY.YY.YY.YY". AWL data from one source will not
affect the other.
-- 
  
  snowjack@fastmail.fm


Re[2]: AWL on SA3.0rc1

Posted by Robert Menschel <Ro...@Menschel.net>.
Hello Daniel,

Saturday, August 21, 2004, 12:19:07 PM, you wrote:

>> (IMO, [AWL] should be off by default to maintain consistency with past
>> versions, but it's not)

DQ> Yeah, that would be more consistent, but my inclination is that we
DQ> should put our best foot forward by default and the AWL *does* improve
DQ> accuracy.  Of course, the AWL algorithm could be better (adding a decay
DQ> is on my list of things to try).

Daniel,

My primary hesitation with AWL has to do with spoofing -- we get too many
spams "from" one of our domain addresses "to" another. I'm concerned that
these will lead AWL to be incorrect wrt our domains. (Since our email
runs on shared web servers, ALL email except webmail comes from external
sources, even those I send to others within our domain.)

Would it be appropriate for me to but in a bugzilla request for something
like the following parameters that would enable us to turn off AWL
processing for specific addresses or domains? If we had that capability,
then I'd have no problem using AWL here.
> awl_exclude_address address-to-exclude
> awl_exclude_domain  domain-to-exclude
and then, since there are bogus addresses that have been harvested and
which are guaranteed to be spam,
> awk_include_address address-to-include
would be beneficial to always flag those as spam.

Bob Menschel




Re: AWL on SA3.0rc1

Posted by Daniel Quinlan <qu...@pathname.com>.
Matt Kettler <mk...@evi-inc.com> writes:

> (IMO, [AWL] should be off by default to maintain consistency with past
> versions, but it's not)

Yeah, that would be more consistent, but my inclination is that we
should put our best foot forward by default and the AWL *does* improve
accuracy.  Of course, the AWL algorithm could be better (adding a decay
is on my list of things to try).

Daniel

-- 
Daniel Quinlan
http://www.pathname.com/~quinlan/

Re: AWL on SA3.0rc1

Posted by Matt Kettler <mk...@evi-inc.com>.
At 10:44 AM 8/20/2004, Dave  Stern - Former Rocket Scientist wrote:
>What is AWL?

The auto-whitelist. (note: this is NOT related to whitelist_from, etc.. 
those are the manual whitelist tools, this is the automatic one. )

>how can I change it?

You can turn it off, as per Theo's suggestion.

(IMO, it should be off by default to maintain consistency with past 
versions, but it's not)


>If it's lying anyway, can I get away with rescoring this rule?
>(header AWL -7.8)

No, it doesn't have any fixed score, it's an averaging system.

Read the fine wiki on this subject, it can tell you what the AWL is, how it 
works, and what things to expect of it.

http://wiki.apache.org/spamassassin/AutoWhitelist