[SLL] [SLL} Re: Anti-spam methods

[Glenn Stone]

On Sun, Jul 29, 2007 at 04:04:32PM -0700, Bill Campbell wrote:
>On Sun, Jul 29, 2007, Glenn Stone wrote:
>>On Sat, Jul 28, 2007 at 05:06:58PM -0400, Nicholas Bodley wrote:
>>Challenge-response kinds of anti-spam don't work very well in corner cases.
>>Consider the case of two people attempting to exchange email, each of which
>>has one of these critters configured.  The first sender elicits a challenge,
>>which runs into his own challenge-response mechanism... *boom*.  Or, even
>>simpler, the case of "I need help, I'm getting on an airplane and I need an
>>answer by the time I land..."
>
>Many serious anti-spammers consider C/R to be abusive, if not
>spam itself because of the blowback produced.
>
>I've found bayesian filters with spamassassin routing messages
>above a user-specified threshhold going into a spam folder for
>review to be very effective in minimizing the effects of spam
>without false positives.  I will drop messages entirely with SA
>scores above some high limit, say 20.0 without fear of loss.

I've found that a well-trained SA implementation can safely kill anything
above 10.0 without fear.  Sometimes I'll get an FP as high as 7.something
with SA, but very very rarely.  

>Spamassassin also allows one to whitelist senders so their
>messages are given a high negative score.

How is this implemented, per-user .whitelistrc or something?  This is very
useful... 

-- Glenn