Bayes - Training "Legitimate" Mail as Spam

[Matt Florido <ma...@floridonet.com>]

I am very new to SA.  I just set it up on Fedora Core 3 last week.  So far
it has been excellent.  I expect performance to even be better once Bayes
establishes itself.

I do have a question with regards to training.  Once in a while, I receive
advertisement e-mails from trusted sites.  What is the best practice to
handle these messages?  Is it more prudent to sa-learn --spam them, or to
simply blacklist?  I say "legitimate" because I do receive mails from
these sources that are of value to me....obviously, I care not for their
advertisement e-mails.

-- 
Regards,
Matt Florido

Re: Bayes - Training "Legitimate" Mail as Spam

[Matt Florido <ma...@floridonet.com>]

Robert Menschel wrote:
> Hello Matt,
> 
> Thursday, February 17, 2005, 6:30:10 AM, you wrote:
> 
[...]
> 
> Why do you receive these advertising emails? What defines a trusted
> site?

As an example, I subscribe my domains through GoDaddy.com.  Once in a 
while, I receive notices such as a domain is going to expire.  I 
consider these important.  More often then not however, I receive 
advertising mails.  For me, these messages are unwanted.  Still, 
GoDaddy.com is what I would classify as "trusted".  You mention another 
good example.

Is it wise to train SA to treat these advertisement mails as spam, or 
will it adversely affect it in some way such as accuracy?  Furthermore, 
can SA be successfully trained to treat one set of mails from a single 
source as Spam, while others are Ham?  Or is best practice simply all 
spam or all ham?

> As with any human endeavor, there's a gray area -- emails which are
> sent to you because of your subscription to something you want, and by
> subscribing you agreed to receive the advertising.

Agreed and unlike typical spam, I don't mind receiving these mails.  I 
just don't want them being sorted with relevant mail.

[...]

> So I accept those ads, and I treat them as non-spam.  I filter/delete
> them in my email client, not in SpamAssassin.  I get one or two of
> these each day. It's my choice, and if they get bothersome, I'll
> unsubscribe.
> 

Point taken.  Perhaps I should reexamine my definition of spam...direct 
advertising from a known company versus broadcast advertising from one 
you're not associated with in any way.

-- 
Regards,
Matt Florido

Re: Bayes - Training "Legitimate" Mail as Spam

[Robert Menschel <Ro...@Menschel.net>]

Hello Matt,

Thursday, February 17, 2005, 6:30:10 AM, you wrote:

MF> I do have a question with regards to training.  Once in a while, I
MF> receive advertisement e-mails from trusted sites.  What is the
MF> best practice to handle these messages?  Is it more prudent to
MF> sa-learn --spam them, or to simply blacklist?  I say "legitimate"
MF> because I do receive mails from these sources that are of value to
MF> me....obviously, I care not for their advertisement e-mails.

Why do you receive these advertising emails? What defines a trusted
site?

As with any human endeavor, there's a gray area -- emails which are
sent to you because of your subscription to something you want, and by
subscribing you agreed to receive the advertising.

Example: I use efax.com's free inbound faxing service. Because of
that, I not only get free inbound faxes emailed to me, but I get
regular advertisements (that supposedly pay for the service). I could
avoid the ads by paying for the service, but I don't get enough faxes
to justify the fees.

In my view this isn't spam -- I could cut off the ads by either paying
for the service or discontinuing the service. I can cut out the ads on
Accuradio.com by not listening to their music. I can cut out the ads
on TV by buying a TIVO or by not watching TV. I can avoid the ads in
the daily newspaper by not reading the newspaper.

So I accept those ads, and I treat them as non-spam.  I filter/delete
them in my email client, not in SpamAssassin.  I get one or two of
these each day. It's my choice, and if they get bothersome, I'll
unsubscribe.

Bob Menschel