Looking for a MoinMoin guru - MoinMoin+SpamBayes == no wiki spam?

[skip]

I have this idea that I should be able to write a security policy for
MoinMoin which uses SpamBayes to judge the appropriateness of any given
submission. It would be more accurate and faster than the current default
security policy in MoinMoin. I started to work on it and have something
that seems to be working on the SpamBayes side of things, but I can't get
the MoinMoin side of the equation to balance.

My messages to the moin-user mailing list have so far fallen on blind eyes,
so I'm broadening my search for a MoinMoin expert who would like to help me
finish off this demonstration. If you're interested in pursuing this idea
here are a couple URLs:

* The current code:

http://spambayes.cvs.sourceforge.net/spambayes/spambayes/spambayes/MoinSecurityPolicy.py?view=log

* My messages to moin-user:

http://article.gmane.org/gmane.comp.web.wiki.moin.general/4381/match=spambayes
http://article.gmane.org/gmane.comp.web.wiki.moin.general/4488/match=spambayes

The motivation for the concept is in the docstring of the module as well as
in the messages, so I won't belabor the point here. As one of the current
"editors" of the Python wiki I can tell you the current scheme of using
lists of "naughty words" to identify potential wiki spam is far from
perfect. There's also the motivation found here:

http://www.publicradio.org/columns/futuretense/

Scroll down and listen to the February 1st episode.

Thanks,

Skip

 

[Fuzzyman]

I have this idea that I should be able to write a security policy for
MoinMoin which uses SpamBayes to judge the appropriateness of any given
submission. It would be more accurate and faster than the current default
security policy in MoinMoin. I started to work on it and have something
that seems to be working on the SpamBayes side of things, but I can't get
the MoinMoin side of the equation to balance.

My messages to the moin-user mailing list have so far fallen on blind eyes,
so I'm broadening my search for a MoinMoin expert who would like to help me
finish off this demonstration. If you're interested in pursuing this idea
here are a couple URLs:

* The current code:

http://spambayes.cvs.sourceforge.net/spambayes/spambayes/spambayes/Mo...

* My messages to moin-user:

http://article.gmane.org/gmane.comp.web.wiki.moin.general/4381/match=...
http://article.gmane.org/gmane.comp.web.wiki.moin.general/4488/match=...

The motivation for the concept is in the docstring of the module as well as
in the messages, so I won't belabor the point here. As one of the current
"editors" of the Python wiki I can tell you the current scheme of using
lists of "naughty words" to identify potential wiki spam is far from
perfect. There's also the motivation found here:

http://www.publicradio.org/columns/futuretense/

Scroll down and listen to the February 1st episode.

You could also try akismet which is designed for finding spam
'comments' in blogs, a similar problem area.

Fuzzyman
http://www.voidspace.org.uk/python/articles.shtml

 

[tleeuwenburg]

I think it's a fantastic idea, myself. I think it should work well.

It's probably important to filter the edit, not the whole article,
otherwise the ham might obscure the spam.

Cheers,
-T

 

[skip]

T> I think it's a fantastic idea, myself. I think it should work well.
T> It's probably important to filter the edit, not the whole article,
T> otherwise the ham might obscure the spam.

That would be a good second phase. My experience with most wiki spam so far
is that the spammers replace the entire content of the page with their junk.
My major concern at the moment is simply to get the MoinMoin mechanism
working. I applied SpamBayes to the form submissions of another website (I
was the author, so I was familiar with the server side code). It worked
great there, so I'm hopeful it will do well in this environment as well.

Skip

 

[skip]

fuzzyman> You could also try akismet which is designed for finding spam
fuzzyman> 'comments' in blogs, a similar problem area.

Thanks. It appears that Akismet is a non-free tool, so doesn't fit in real
well with the open source nature of MoinMoin and SpamBayes. I think I'll
leave it to someone else to implement a security policy using that. It does
seem like it would be trivial to write such a policy. The
revision/reversion mechanism in MoinMoin is what's giving me fits. Akismet
clearly sidesteps that problem altogether.

Skip