You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Raquel Rice <ra...@thericehouse.net> on 2004/02/14 04:15:07 UTC
Bayes Filtering (was Re: Some real anti-bayes stuffing )
On Fri, 13 Feb 2004 16:05:02 -0800
jm@jmason.org (Justin Mason) wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> Mark A. DeMichele writes:
> > This is exactly what I was talking about in a previous post.
> >
> > If spammers start doing this and they make sure this bogus
> > section is larger than the actual spam section, the bayes filter
> > will probably mark it as ham. What's worst is if you then force
> > the bayes filter to learn this as spam, now you just increased
> > the spam score for each of these good words. If this happens
> > over and over again, I would imagine that the bayes filter would
> > malfunction. At least that's my opinion, but feel free to
> > disagree.
>
> Good, will do.
>
> The idea of bayes is that you train it on
>
> 1. *YOUR* ham
> 2. *YOUR* spam
>
> Unless spammers figure out what *YOU* call ham, they can add
> random words, bits of Russian literature, snippets of Tom Sawyer
> until the cows come home.
>
> For spammers to effectively "poison" bayes, they need to figure
> out what kind of text *YOU* have trained on. If I don't receive
> copies of Tom Sawyer by email normally, then sure, 19th-century US
> lit will become a spam sign.
>
> But I don't care because *I DON'T* receive copies of Tom Sawyer by
> email, normally. (I reserve that honour for snailmail, or
> occasionally by FTP.) So it's not going to wind up misclassifying
> anything as a result.
>
> In the worst case, they'll find one or two strong ham-sign words
> -- like'Kits', or 'entries' (for my corpus). Worst case? I
> retrain on their mail, and those tokens become about even ham and
> spam counts, 0.5 probability, and are *ignored* by the Bayes
> calculation in future.
>
> *PLEASE* read up on how Bayes works. READ John Graham-Cumming's
> presentation from the last Spam Conf, and NOTICE how it took him
> thousands of iterations of bayes-poisoning, sending a mail each
> time with a direct feedback loop, to get a single spam through.
>
> The sky is NOT falling, guys!
>
> - --j.
Justin,
Thank you for a good explanation of how the bayes filters work. It
seems to me that the onus is on me then to make sure whatever mail I
get, gets fed back to the "bayes learning engine". That way by the
time the flood hits, I've trained our filters to respond correctly.
--
Raquel
============================================================
Never fear shadows. They simply mean there's a light shining
somewhere nearby.
--Ruth E. Renkel
Re: Bayes Filtering (was Re: Some real anti-bayes stuffing )
Posted by Lucas Albers <ad...@cs.montana.edu>.
Raquel Rice said:
> Thank you for a good explanation of how the bayes filters work. It
> seems to me that the onus is on me then to make sure whatever mail I
> get, gets fed back to the "bayes learning engine". That way by the
> time the flood hits, I've trained our filters to respond correctly.
>
Just set your autolearn spam/ham high and low enough respectivelly that
you don't have to teach it.
I've handled a couple of thousand messages in the past few months, and I
haven't had to retrain it either way.
I'm lazy, and my users have not complained, and my thresholds are high
enough, that while it might not catch all the spam, it really never
classifies something as wrong.
I set autolearn_ham to from .00 to .4
and
and autolearn_spam from 9.5 to 12
Depending teh variety of mail I receive at the server, and whether it is
an internal or or external server.
--
Luke Computer Science System Administrator
Security Administrator,College of Engineering
Montana State University-Bozeman,Montana