SitePoint.com

ArticleBot · Nov 23, 2003, 19:44

This is a dedicated thread for discussing the SitePoint article 'Review - Spamnix with Bayesian Filter'

00.ZoMbiE · Nov 23, 2003, 19:45

For outlook, you can get a free spam filter with Bayesian Filter like http://starship.python.net/crew/mhammond/spambayes/ . In terms of functionality, it works like Spamnix but it's free. I've used it for a couple of weeks now and it has worked great for me.

Nov 23, 2003, 23:58

Article was very good.
But if could give the details about how actually bayesian filter works it would be really helpful.

Nov 24, 2003, 03:51

Mozilla comes with Bayesian filters builtin ... and has for quite a long time, actually. It also comes without the plethora of security problems found in MS products like Outlook Express. Not to mention that it is more customizable, more closely standards compliant, and best of all, it doesn't cost a dime ...

Nov 24, 2003, 07:03

Have your mail provider install SpamAssassin, it includes a Bayes learning machine among many other spam tests (DNS blacklists)

spybreak · Nov 24, 2003, 07:51

Article: A Plan for Spam

I successfully use this Bayesian Spam Filter (Windows): K9
a) looks good
b) gives the impression that it's very well coded
c) works very well for me
d) works with all pop3 programs

Nov 24, 2003, 08:18

Basically, a Bayesean filter works by ranking the words found in your email. Words that are found in spam are given a high spam ranking. Words found in non-spam are given low spam rankings. For example, seeing "viagra" in an e-mail may indicated spam 99% of the time. "Compiler" may indicate a low probability of spam. From these rankings, the filter looks at a new e-mail, looks at the highest- and lowest-ranked words, and generates an overall probability that the e-mail is spam. One of the really great things about this filter is that it looks at more than the text of the e-mail: it will also look at message headers and check hypertext (a lot of spam is caught because many use red text). In addition, the filter becomes very personal. "Insurance" may indicate spam for many people, but in an insurance office, it will indicate a lower probability, and that office's filter will have different spam triggers. This makes it very difficult for spammers to get around the filter, since what works for one person may not work for another. And since the filter learns to catch any uncaught spam, these new spam messages will be caught soon, anyway.
My office has been using the Python Outlook SpamBayes plug-in for about six months...and now I can't ever see a time when we don't have a Bayesean spam filter.

Nov 25, 2003, 10:04

I use Thunderbird e-mail client which is available for free at http://www.mozilla.org/projects/thunderbird/ that has Bayesian spam filtering built in and it has been working really well for e. Thunderbird is part of Open source project Mozilla

veronicay · Nov 27, 2003, 01:36

The latest paid-for version of Eudora has Bayesian filtering built in, so you don't need to bother with a separate product. It's completely seamless and works great -- from day 1, 98% of my spam ended up in the "Junk" folder and now accuracy is around 99%, with virtually no false positives. Very worthwhile if you get lots of spam.

Nov 27, 2003, 17:35

I have also had experienced abnormal rates of spam per day and found my solution to be a open source program called 'POPFile'.

It acts as a proxy daemon between your POP3 aware e-mail client and the actual POP3 server, using this 'man-in-the-middle' concept, it is able to build up a profile of what is spam and what is not and insert headers into the e-mail which you can then later detect using your e-mail client, I currently using 'spam', 'newsletters', 'inbox', 'ebay', and 'paypal'.

It has so far done remarkably well, and stays at a constant success rate of around 97.83%, though this was much lower when the program was learning from scratch.

I recommend it if anyone wishes to try it as their spam solution. The URL is:
http://popfile.sf.net/

Nov 29, 2003, 00:40

do a search for k9... works really well... and the smallest footprint off all spam filters...

Bearwalk · Dec 4, 2003, 02:02

For users of Opera's M2, there's a similar feature in production:
http://my.opera.com/forums/showthrea...threadid=33875

It isn't Bayesian, but works great so far! I fed it with 3000 non-spams and 13000 spams (I've been saving them for a year), and the only footprint was a 200kB index file.

Nov 23, 2003, 19:44	#1
ArticleBot SitePoint Articles Join Date: Apr 2001 Posts: 0	Discussion thread for Review - Spamnix with Bayesian Filter This is a dedicated thread for discussing the SitePoint article 'Review - Spamnix with Bayesian Filter'

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Nov 23, 2003, 19:45	#2
00.ZoMbiE SitePoint Zealot Join Date: Jul 2003 Location: Metro Manila, Philippines Posts: 173	For outlook, you can get a free spam filter with Bayesian Filter like http://starship.python.net/crew/mhammond/spambayes/ . In terms of functionality, it works like Spamnix but it's free. I've used it for a couple of weeks now and it has worked great for me.

Nov 23, 2003, 23:58	#3
Anonymous SitePoint Community Guest Posts: n/a	Article was very good. But if could give the details about how actually bayesian filter works it would be really helpful.

Nov 24, 2003, 03:51	#4
Anonymous SitePoint Community Guest Posts: n/a	Mozilla comes with Bayesian filters builtin ... and has for quite a long time, actually. It also comes without the plethora of security problems found in MS products like Outlook Express. Not to mention that it is more customizable, more closely standards compliant, and best of all, it doesn't cost a dime ...

Nov 24, 2003, 07:03	#5
Anonymous SitePoint Community Guest Posts: n/a	Have your mail provider install SpamAssassin, it includes a Bayes learning machine among many other spam tests (DNS blacklists)

Nov 24, 2003, 07:51	#6
spybreak SitePoint Addict Join Date: Apr 2003 Location: Germany Posts: 230	Article: A Plan for Spam I successfully use this Bayesian Spam Filter (Windows): K9 a) looks good b) gives the impression that it's very well coded c) works very well for me d) works with all pop3 programs

Nov 24, 2003, 08:18	#7
Anonymous SitePoint Community Guest Posts: n/a	Basically, a Bayesean filter works by ranking the words found in your email. Words that are found in spam are given a high spam ranking. Words found in non-spam are given low spam rankings. For example, seeing "viagra" in an e-mail may indicated spam 99% of the time. "Compiler" may indicate a low probability of spam. From these rankings, the filter looks at a new e-mail, looks at the highest- and lowest-ranked words, and generates an overall probability that the e-mail is spam. One of the really great things about this filter is that it looks at more than the text of the e-mail: it will also look at message headers and check hypertext (a lot of spam is caught because many use red text). In addition, the filter becomes very personal. "Insurance" may indicate spam for many people, but in an insurance office, it will indicate a lower probability, and that office's filter will have different spam triggers. This makes it very difficult for spammers to get around the filter, since what works for one person may not work for another. And since the filter learns to catch any uncaught spam, these new spam messages will be caught soon, anyway. My office has been using the Python Outlook SpamBayes plug-in for about six months...and now I can't ever see a time when we don't have a Bayesean spam filter.

Nov 25, 2003, 10:04	#8
Anonymous SitePoint Community Guest Posts: n/a	I use Thunderbird e-mail client which is available for free at http://www.mozilla.org/projects/thunderbird/ that has Bayesian spam filtering built in and it has been working really well for e. Thunderbird is part of Open source project Mozilla

Nov 27, 2003, 01:36	#9
veronicay SitePoint Zealot Join Date: Oct 1999 Location: France Posts: 144	The latest paid-for version of Eudora has Bayesian filtering built in, so you don't need to bother with a separate product. It's completely seamless and works great -- from day 1, 98% of my spam ended up in the "Junk" folder and now accuracy is around 99%, with virtually no false positives. Very worthwhile if you get lots of spam.

Nov 27, 2003, 17:35	#10
Anonymous SitePoint Community Guest Posts: n/a	I have also had experienced abnormal rates of spam per day and found my solution to be a open source program called 'POPFile'. It acts as a proxy daemon between your POP3 aware e-mail client and the actual POP3 server, using this 'man-in-the-middle' concept, it is able to build up a profile of what is spam and what is not and insert headers into the e-mail which you can then later detect using your e-mail client, I currently using 'spam', 'newsletters', 'inbox', 'ebay', and 'paypal'. It has so far done remarkably well, and stays at a constant success rate of around 97.83%, though this was much lower when the program was learning from scratch. I recommend it if anyone wishes to try it as their spam solution. The URL is: http://popfile.sf.net/

Nov 29, 2003, 00:40	#11
Anonymous SitePoint Community Guest Posts: n/a	do a search for k9... works really well... and the smallest footprint off all spam filters...

Dec 4, 2003, 02:02	#12
Bearwalk SitePoint Member Join Date: Dec 2003 Location: Kalmar, Sweden Posts: 2	For users of Opera's M2, there's a similar feature in production: http://my.opera.com/forums/showthrea...threadid=33875 It isn't Bayesian, but works great so far! I fed it with 3000 non-spams and 13000 spams (I've been saving them for a year), and the only footprint was a 200kB index file.