about traing bayes method.

[Monty Ree <ch...@hotmail.com>]

Hello, list.

I would like to ask some about bayes.

 If I have recevied ham mail which written ***SPAM***, 
   So in order to train bayes this mail with sa-learn to ham, I forwarded 
this mail using outlook or outlook express to ham@test.com and this mail 
would be saved at /var/spool/mail/ham.

  and if I execute like this, this would be meaningful or meaningless?

 # sa-learn --ham /var/spool/mail/ham

 If meaningless, Any good solution which like this?

 I have operated several domains at mail servers.So I can't train all 
mails.


Thanks for your time...

_________________________________________________________________
메신저에서 문자를 바로 보내보세요 http://phonebuddy.msn.co.kr/ 

Re: about traing bayes method.

[Ian <sp...@FishNet.co.uk>]

On 1 Feb 2007 at 11:24, maillist wrote:

<Snip>

> I think one quick and easy way to accomplish this is to make your
> outlook clients leave a copy on the server. When they report a spam
> message, and you want to learn it as such, go to the server, and learn
> it either from their inbox directly, or copy it "as is" to another file,
> and learn that one. This is by no means a long-term way to deal with
> this, but it's what I did/do.

Hi,

Sorry for jumping on this one a bit late.  This method may help if you have exchange:

http://wiki.mailscanner.info/doku.php?id=documentation:anti_spam:spamassassin:sa-
learn:msexchange

It describes how to set up a public folder on Exchange that your users can move spam 
messages to. A python script on the server downloads these messages periodically and 
runs sa-learn on them.  It was designed for MailScanner but all you would have to do is 
change the PREFS file location.

I have never tested this so no idea if it works.

Regards

Ian
-- 

Re: about traing bayes method.

[maillist <ma...@emailacs.com>]

Matt Kettler wrote:
> Monty Ree wrote:
>   
>> Hello, list.
>>
>> I would like to ask some about bayes.
>>
>> If I have recevied ham mail which written ***SPAM***, So in order to
>> train bayes this mail with sa-learn to ham, I forwarded this mail
>> using outlook or outlook express to ham@test.com and this mail would
>> be saved at /var/spool/mail/ham.
>>
>> and if I execute like this, this would be meaningful or meaningless?
>>     
> Meaningless. Forwarding the message created a new, completely different,
> message, at least from a bayes perspective. It has new headers, and all
> the mime sections were likely reformatted and re-encoded to outlook's
> liking. While there's a lot of visual similarity when rendered by
> outlook, the raw message is quite different.
>   
>> # sa-learn --ham /var/spool/mail/ham
>>
>> If meaningless, Any good solution which like this?
>>     
> I'm no outlook expert, but what you're looking for is a way to extract
> the original message, with its original headers and mime sections.
>
>   
I think one quick and easy way to accomplish this is to make your
outlook clients leave a copy on the server. When they report a spam
message, and you want to learn it as such, go to the server, and learn
it either from their inbox directly, or copy it "as is" to another file,
and learn that one. This is by no means a long-term way to deal with
this, but it's what I did/do.

-=Aubrey=-

Re: about traing bayes method.

[Matt Kettler <mk...@verizon.net>]

Monty Ree wrote:
> Hello, list.
>
> I would like to ask some about bayes.
>
> If I have recevied ham mail which written ***SPAM***, So in order to
> train bayes this mail with sa-learn to ham, I forwarded this mail
> using outlook or outlook express to ham@test.com and this mail would
> be saved at /var/spool/mail/ham.
>
> and if I execute like this, this would be meaningful or meaningless?
Meaningless. Forwarding the message created a new, completely different,
message, at least from a bayes perspective. It has new headers, and all
the mime sections were likely reformatted and re-encoded to outlook's
liking. While there's a lot of visual similarity when rendered by
outlook, the raw message is quite different.
>
> # sa-learn --ham /var/spool/mail/ham
>
> If meaningless, Any good solution which like this?
I'm no outlook expert, but what you're looking for is a way to extract
the original message, with its original headers and mime sections.

Re: about traing bayes method.

["Chris St. Pierre" <st...@NebrWesleyan.edu>]

On Thu, 1 Feb 2007, Monty Ree wrote:

>  So in order to train bayes this mail with sa-learn to ham, I forwarded this 
> mail using outlook or outlook express to ham@test.com and this mail would be 
> saved at /var/spool/mail/ham.
>
> and if I execute like this, this would be meaningful or meaningless?
>
> # sa-learn --ham /var/spool/mail/ham

We do something similar (although a bit more involved).  You're on the
right track, but a few tweaks might help:

First, you'll want to resend the message, not forward it; that ensures
that the message is kept intact.

Second, it looks like you're delivering to Mbox-style mail boxes, so
you'll want to do:

$ sa-learn --mbox --ham /var/spool/mail/ham

Hope that helps!

Chris St. Pierre
Unix Systems Administrator
Nebraska Wesleyan University
----------------------------
Never send mail to thobrux@nebrwesleyan.edu