Re[2]: [Bug 891] New: look at integrating external spam detection library

On Wed, 21 Apr 2004 12:10:48 +1000 Mark Livingstone <[email protected]> wrote:

ML> > Either we should rely on external (ISP-based for example) spam filters or we
ML> > should integrate an existing spam detection library. One possible condidate is
ML> > http://www.nuclearelephant.com/projects/dspam/
ML> 
ML> I spend some time in anti-spam circles and I have never heard of anyone
ML> using it. I also notice that a prerequisite for using it is a unix
ML> based mail server. This would suggest it is going to be a server based
ML> solution rather than a client based one (or is that what you intended)?

 No, it consists of 2 components: the one you saw (working on mail server)
and a library. I'm interested in using the library.

ML> The site appears to me (at first glance) to be full of buzz words but
ML> somewhat short on substance.

 I don't know much about it but it was recommended to me by someone whom I
respect so I'd like to at least try it out...

ML> > If you know of others, please add them here.
ML> 
ML> popfile@sourceforge PERL
ML> spambayes@sourceforge Python
ML> 
ML> Both can be used as is or further integrated into mail packages. Both
ML> inherently platform independant. 

 I think we need something in C/C++.

ML> I have used Popfile for years and can attest to it's efficacy. 99.85%+
ML> success is very achievable. It uses the Train on Error system so you
ML> only classify messages it got wrong reducing corpus size and reducing
ML> corpus poisoning attempts by spammers et al. It is also aimed at being
ML> a text classification system more than just a spam detector and with
ML> very little work could replace most of the filtering capabilities found
ML> in Mahogany at present. 

 But we'd need to always embed Perl then... I'm not sure I like it.

ML> It can also be very successfully trained to sort out the spam and the
ML> ham from a particular address. Case in point, I am on the wxPython
ML> mailing list. It is often spammed, gets those "unsubscribe" messages
ML> in-list, gets malware type messages, gets on-vacation annoucements from
ML> list users, bounced messages, "I found a virus" messages from
ML> anti-virus scanners etc yet I do not see those because I have trained
ML> it to discriminate them in multiple languages too!

 This does look good. Of course, if I understand it correctly, dpsam
pretends to do the same.

ML> I have not used Spambayes but have heard of good experiences with it. 

 Is SpamAssasin engine available as a library?

 Thanks,
VZ



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click