Re: Pyzor fingerprinting approach.

[Jesper K. <je...@kr...>]

Frank Tobin sagde:
> Jesper Krogh, on 2004-04-19, wrote:
>
>> I seem to get more and more email with a rather large random generated
>> section of words ( not sentences ). Does Pyzor have any chance agains
>> them?
>
> It certainly depends.  One would hope that messages that are 99%
> randomness make for poor sales, so that would be ineffective spam, and
> self-defeating in that sense.  Bayes-like systems would be good against
> those types of spam, I would suspect.

I don't really think that the bayes systems are good.. it's statistics
afterall... most of the ham-mail i recieve is Danish and most of the
spam-mail is English so I actually train my bayes filter to recognise
Danish vs. English instead of Spam vs. Ham.... I suspect.

Example:
http://krogh.cc/spammail.txt

>  The key thing to remember is that
> Pyzor is only one part of an anti-spam solution.

Sure ... but I believe that these fingerprinting approaches do quite well..

Seems like razor' is more or less dead as addon to spamassassin so there
is pyzor left. Sometimes I get Timeout errors tring to run pyzor report
66.250.40.33:24441      TimeoutError:

Would anyone mind that I try to get this list/group on Gmane?

-- 
Jesper Krogh