[unisog] Anyone have procmail script for sanity checking submissions to spam assassin bayesian stuff?

Russell Fulton r.fulton at auckland.ac.nz
Tue Mar 11 02:45:48 GMT 2008


On 11/03/2008, at 10:59 AM, Valdis.Kletnieks at vt.edu wrote:

> On Tue, 11 Mar 2008 09:48:16 +1300, Russell Fulton said:
>> Currently we eyeball all the missed spam that we feed back to the
>> Bayesian engine in spam assassin because we get folk forwarding stuff
>> from the VC and departmental mailling lists.  We also need to make
>> sure that submissions have full headers etc.
>>
>> This is becoming seriously onerous and needs to be automated.
>
> How do you automate what's probably the most important part - double- 
> check
> that something actually *is* spam, and somebody didn't miscue and  
> hit that
> ooh-shiny "report as spam" button by accident?

we feed email originating from reliable hosts on campus to the  
Bayesian engine as ham.  We encourage people to forward missed spam to  
a special email address where we check it before feeding it to spam  
assassin (so we don't have the AOL problem where (I'm told) the mark  
as spam button is right next to the delete button.

I too have my doubts as to exactly how effective this is particularly  
given that most of the stuff getting through now is explicitly  
designed to confuse a Bayesian filter  (over 90% random garbage) .    
I'm on the point of saying, if we can't automate this then we skip the  
Bayesian filter altogether.  Hence my original question.

Russell




More information about the unisog mailing list