Bayes corrupt?

[Jeff Chan <je...@surbl.org>]

Hi All,
One of our SA 3.1.0 installations is giving BAYES_00=-2.599 for
this definite spam.  Does that mean our Bayes database is corrupt?

> X-Spam-Status: No, score=0.863 tagged_above=0 required=2
>  tests=[BAYES_00=-2.599, DATE_IN_PAST_06_12=0.827, FORGED_RCVD_HELO=0.135,
>  RAZOR2_CF_RANGE_51_100=0.5, RAZOR2_CF_RANGE_E4_51_100=1.5, RAZOR2_CHECK=0.5]

> Dear Marry, 
> 
> A while back I was  let go from a job I held for 312 months would you
> believe. Its very hard to thank you enough for launching  me in this
> industry. You have given me a exciting beginning on life.  Already making
> twice as much as I took home in my old job.  
> 
> 
> I purchased for cash a new Jaguar. Taking home 165,000US in 18 months.
> Having a great time in this profession. It is fun and I am a hero to the
> courts and to my customers. What an exciting job to be in.
> 
> Following exactly what your instruction tells me to do, is proceeding
> perfectly.  I go to the local court and locate all of the customers I can
> handle. 
> 
> 
> I utilize your advanced reporting services to find all items which can be
> seized. Using your fill in the blank forms I deliver them to the appropriate
> firms. Then the funds arrive to my PO Box.  Its like magic.  Its so exciting
> opening up the payments as they arrive.. 
> 
> I can take a holiday when ever we have a notion to do so.  England and
> Japan this year. 
> 
> Provide this letter to others.  This profession is so untapped it needs
> many more of us assisting the courts and the  people who have been harmed.
> 
> Sincerely,
> Tara W.    WA    
> 
> 
> This can be you! 
> 
> Tele-Phone
> 1- 303--395--3900
> Higher to learn more or to end receiving more info and then to see location
> 
> The two shipwrecked sailors were still sound asleep. Rob stood up and
> waited for the strangers to draw nearb58



Jeff C.
-- 
Jeff Chan
mailto:jeffc@surbl.org
http://www.surbl.org/

Re: Bayes corrupt?

[Loren Wilton <lw...@earthlink.net>]

> One of our SA 3.1.0 installations is giving BAYES_00=-2.599 for
> this definite spam.  Does that mean our Bayes database is corrupt?

I'd say that it is at the very least mis-trained as far as this spam is
concerned.  Without further evidence I'd hesitate to state outright that it
is corrupt.  If this is common for many spams I might be inclined to
consider that it has in some way become totally mistrained or outright
corrupted.

The usual sign of a bayes database gone south is giving bayes_99 to
everything.  Giving bayes_00 to everything would imply that it had somehow
managed to purge most of the spammy tokens while retaining at least a few
ham tokens.  Since spam usually outnumbers ham, this would be a bit strange
if you have autolearning enabled.  It might be possible if you are manually
training and haven't trained recently.

        Loren