Re: [gentoo-user] Re: [lists] [gentoo-user] Bayesian spam filtering
On 2003-06-05 07:15, Anthony Ventimiglia <[EMAIL PROTECTED]> wrote: > Bogofilter should be more *nixy since it was written by esr, but I > don't know if it supports more than two buckets. Not sure what you mean by buckets. Bogofilter doesn't really care how many mailboxes you use. It operates on one message at a time. Then you filter on the X-Bogosity header. If you mean classification buckets, Bogofilter can be set to classify mail as spam, not spam (ham), and not sure. That's three. -- Luke -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
If you are using qmail, consdier qconfirm: It requires you to expose your qmail to the net but as long as you have an always on connection, it is easy to do with dyndns.org. It virtual elminates SPAM, because only those that have explicitly responded to a confirmation request, or you have explicitly placed in a whitelist, are allowed to delivered to your inbox. You can find an ebuild compressed tarball here: http://bugs.gentoo.org/attachment.cgi?id=12442&action=view Put it in a PORTAGE_OVERLAY_DIR and just emerge. Lincoln On Thu, 2003-06-05 at 09:31, Larry Wright wrote: > I am currently running qmail + fetchmail + procmail + spamassassin. Everything > works pretty well, but I'm a little dissapointed in spamassassin's accuracy. > I'd ideally like something similar to spambayes, which is trainable. > Unfortunately spambayes does not appear to work with maildir. I'd really like > to keep qmail, does anyone know of a bayesian filter that works with > qmail/maildir? > > Thanks, > Larry > > > -- > [EMAIL PROTECTED] mailing list > -- Lincoln A. Baxter <[EMAIL PROTECTED]> -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
On 2003.06.05 09:31, Larry Wright wrote: I am currently running qmail + fetchmail + procmail + spamassassin. Everything works pretty well, but I'm a little dissapointed in spamassassin's accuracy. I'd ideally like something similar to spambayes, which is trainable. Unfortunately spambayes does not appear to work with maildir. I'd really like to keep qmail, does anyone know of a bayesian filter that works with qmail/maildir? Just to tag a "me too" on here. I'm using Spam Assassin and rarely get any spam in my inbox. I am very impressed with this product. As for maildir support, they way I set up SA with procmail negates SA's need to even touch maildirs. I'd assume a similar setup is possible for any spam filter. -Chris I I bought some used paint. It was in the shape of a house. -- Steven Wright pgp0.pgp Description: PGP signature
Re: [gentoo-user] Bayesian spam filtering
On Thu, 2003-06-05 at 09:40, Ryan wrote: > I've been using bogofilter (http://bogofilter.sourceforge.net/) with > procmail with great success. A few false negatives now and then but I > haven't seen a false positive yet. The only thing I don't like about the > setup is that I have to go to a commandline to mark the spam that slipped > through bogofilter as spam. I wish that I could just throw something in a > cron job that every couple hours it would re-compute its wordlists based > on the contents of my spam folders and my other folders. Does anyone have > a script that does this? I hear that spamprobe > (http://spamprobe.sourceforge.net/) (not used it yet) has this capability > built-in. Maybe I'll try switching to see how it works. I've been using spamprobe for months now. It is wonderful. Move spam to spam mbox and spamprobe does the rest. You can find an ebuild here: http://bugs.gentoo.org/show_bug.cgi?id=19192 Once installed, I use the following cron entries: spamprobe good /home/skk/mail/inbox spamprobe spam /home/skk/mail/spam and the procmail entry: :0 SCORE=| /usr/bin/spamprobe receive :0 wf | formail -I "X-SpamProbe: $SCORE" :0 a: *^X-SpamProbe: SPAM spam I'm using mbox, I have not used this is with maildir, or other mail storage methods. -- Steven Knight [EMAIL PROTECTED] IM : skkataim and tho' We are not now that strength which in old days Moved earth and heaven, that which we are, we are,-- One equal temper of heroic hearts, Made weak by time and fate, but strong in will To strive, to seek, to find, and not to yield. -- Ulysses by Alfred Lord Tennyson -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
On Thu, Jun 05, 2003 at 08:31:48AM -0500 or thereabouts, Larry Wright wrote: > I am currently running qmail + fetchmail + procmail + spamassassin. Everything > works pretty well, but I'm a little dissapointed in spamassassin's accuracy. As Jens said, the latest version of SpamAssassin supports bayesian filtering. I've been using it for almost 2 weeks now and, after taking the time to train it properly, it has been remarkably accurate. --kurt pgp0.pgp Description: PGP signature
Re: [gentoo-user] Bayesian spam filtering
* On Thu, Jun 05, 2003 at 08:31:48 -0500, Larry Wright wrote: > I am currently running qmail + fetchmail + procmail + spamassassin. Everything > works pretty well, but I'm a little dissapointed in spamassassin's accuracy. > I'd ideally like something similar to spambayes, which is trainable. > Unfortunately spambayes does not appear to work with maildir. Try to use the latest SpamAssassin (2.55), which supports Maildirs (you can train your Bayesian filter by 'sa-learn --dir --ham Maildir/.foobar/cur' resp. 'sa-learn --dir --spam...'). Read http://au.spamassassin.org/doc/sa-learn.html before starting to train your filters. Using SpamAssassin, new mail above and beyond a specified score (defaults: +15 and -2, IIRC) will train SpamAssassin's bayesian filter automagically with ham/spam. After a certain amount of learned mails, the bayes-filter will start to work. I'm using this setup on a Debian box, installing and updating SpamAssassin via CPAN. Regards, Jens -- BOFH Excuse #420: Feature was not beta tested -- [EMAIL PROTECTED] mailing list
[gentoo-user] Re: [lists] [gentoo-user] Bayesian spam filtering
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Thursday 05 June 2003 09:31 am, Larry Wright wrote: > I am currently running qmail + fetchmail + procmail + spamassassin. > Everything works pretty well, but I'm a little dissapointed in > spamassassin's accuracy. I'd ideally like something similar to spambayes, > which is trainable. Unfortunately spambayes does not appear to work with > maildir. I'd really like to keep qmail, does anyone know of a bayesian > filter that works with qmail/maildir? > Drop that Spamassassin, it's tacking Bayesian filters on in hopes to stay alive, because Bayesian is the only way to do it. I just installed Gentoo a couple weeks ago, on my last box I'd been using my own Bayesian filter which basically worked via procmail, mine was kind of crude, so I installed popfile (popfile.sf.net) when I set up Gentoo. Popfile has a real nice HTTP based control interface, but it is unfortunatley does not do things the Linux way. Bogofilter should be more *nixy since it was written by esr, but I don't know if it supports more than two buckets. -BEGIN PGP SIGNATURE- Version: GnuPG v1.2.2 (GNU/Linux) iQCVAwUBPt9O7gqNYTLzAsoIAQLtwQQAno2mVWR4LFyRF3OaePqKjkQ9EjBxd366 bgS9Yh8Gbj2T+xy7TRzsJ7FDlGxh4otmfRhYfOH1GXYaQzrBNLGyZCwJuL8PGnXI FnhykI+PlYxa8Mm5TRraJH49h7VD0+yTr11pFNwYACM1EiMu0TXiUX3mC3NewwtW JVJrftC/yzY= =C/Ol -END PGP SIGNATURE- -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
Try the current spamassassin (2.54). ACCEPT_KEYWORDS=~x86 emerge dev-perl/Mail-SpamAssassin Looking for bayesian only then bogofilter may be a better choice (I prefer spamassassin on the server & popfile on the client). emerge net-mail/bogofilter Paul At 08:31 AM 06/05/2003 -0500, you wrote: I am currently running qmail + fetchmail + procmail + spamassassin. Everything works pretty well, but I'm a little dissapointed in spamassassin's accuracy. I'd ideally like something similar to spambayes, which is trainable. Unfortunately spambayes does not appear to work with maildir. I'd really like to keep qmail, does anyone know of a bayesian filter that works with qmail/maildir? Thanks, Larry -- [EMAIL PROTECTED] mailing list -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
I've been using bogofilter (http://bogofilter.sourceforge.net/) with procmail with great success. A few false negatives now and then but I haven't seen a false positive yet. The only thing I don't like about the setup is that I have to go to a commandline to mark the spam that slipped through bogofilter as spam. I wish that I could just throw something in a cron job that every couple hours it would re-compute its wordlists based on the contents of my spam folders and my other folders. Does anyone have a script that does this? I hear that spamprobe (http://spamprobe.sourceforge.net/) (not used it yet) has this capability built-in. Maybe I'll try switching to see how it works. Ryan [EMAIL PROTECTED] > I am currently running qmail + fetchmail + procmail + spamassassin. > Everything > works pretty well, but I'm a little dissapointed in spamassassin's > accuracy. > I'd ideally like something similar to spambayes, which is trainable. > Unfortunately spambayes does not appear to work with maildir. I'd really > like > to keep qmail, does anyone know of a bayesian filter that works with > qmail/maildir? > > Thanks, > Larry > > > -- > [EMAIL PROTECTED] mailing list > > -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
bogofilter might be what you're looking for. I'm using it on my postfix+mailfilter system, but it should work fine with procmail. Wes On Thu, 5 Jun 2003, Larry Wright wrote: > I am currently running qmail + fetchmail + procmail + spamassassin. Everything > works pretty well, but I'm a little dissapointed in spamassassin's accuracy. > I'd ideally like something similar to spambayes, which is trainable. > Unfortunately spambayes does not appear to work with maildir. I'd really like > to keep qmail, does anyone know of a bayesian filter that works with > qmail/maildir? > > Thanks, > Larry > > > -- > [EMAIL PROTECTED] mailing list > > -- [EMAIL PROTECTED] mailing list
[gentoo-user] Bayesian spam filtering
I am currently running qmail + fetchmail + procmail + spamassassin. Everything works pretty well, but I'm a little dissapointed in spamassassin's accuracy. I'd ideally like something similar to spambayes, which is trainable. Unfortunately spambayes does not appear to work with maildir. I'd really like to keep qmail, does anyone know of a bayesian filter that works with qmail/maildir? Thanks, Larry -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Friday 06 June 2003 01:09 am, Luke Ravitch wrote: > On 2003-06-05 20:45, Anthony Ventimiglia <[EMAIL PROTECTED]> wrote: > > That's what I mean by buckets, Popfile lets you have as many as you want, > > which I like so I can sort normal mail, work related, mailing lists, > > spam, etc.. as many categories as I want. > > Well, you were right about bogofilter being "more *nixy" ;-) > > Normally I would use procmail rules for sorting based on mailing > lists, work related, etc. On the other hand, doing all that sorting > based on Bayesian filtering sounds pretty cool. And, for that matter, > such a capability probably wouldn't make bogofilter any less *nixy. > It would still be doing one thing (Bayesian filtering) and (hopefully) > doing it well. > It's a fairly simple thing to enable it to work with more "buckets" the Bayesian formula allows for any amount. When I first read about Bayesian filters, I wrote my own fairly quickly, and I actually have an almost complete filter that allows for any amount of 'buckets', I just recently set up this Gentoo system, so I installed Popfile so I didn't have to install my library and filter. I really like the popfile interface, so I'll probably end up doing something similar with my own filter, once I get back to finishing it. > I guess simple things like To and From headers work well enough for > mailing lists (and I would normally keep a separate address for work > vs. home) that I hadn't even thought of using Bayesian filtering for > anything but spam. Only with spammers can I not trust the simple > things! > > Spammers should [fate far worse than anything I can think of]! Anyone who hasn't looked at Bayesian filters should, They give excellent results with a very simple algorithm, and all you have to do is train it when you start out. Using more categories makes it learn a little slower, but usually after seeng just a handful, a Bayesian filter already predicts with about 80% accuracy. The one I have setup now has 5 categories and is at 94% accuracy after about 1200 messages. -BEGIN PGP SIGNATURE- Version: GnuPG v1.2.2 (GNU/Linux) iQCVAwUBPuAovwqNYTLzAsoIAQI8mQP/W/rifRBlCI+ltlZ/Vqi9iSU/VlmnWdyP vSjEYo1UJg7e5XxN3EQDFCsEDGLMD+Y6WeOamQ70tvX4VL39mAWW+pEF1+s5lOuM MWGyKB0Jilc4y1Nk+L2t9XrcUuDqHXaUoLwWoKAyndGfF4wwQug61ILcMmu/8rRn B33625ElVPA= =RztF -END PGP SIGNATURE- -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Bayesian spam filtering
On 2003-06-05 20:45, Anthony Ventimiglia <[EMAIL PROTECTED]> wrote: > That's what I mean by buckets, Popfile lets you have as many as you want, > which I like so I can sort normal mail, work related, mailing lists, spam, > etc.. as many categories as I want. Well, you were right about bogofilter being "more *nixy" ;-) Normally I would use procmail rules for sorting based on mailing lists, work related, etc. On the other hand, doing all that sorting based on Bayesian filtering sounds pretty cool. And, for that matter, such a capability probably wouldn't make bogofilter any less *nixy. It would still be doing one thing (Bayesian filtering) and (hopefully) doing it well. I guess simple things like To and From headers work well enough for mailing lists (and I would normally keep a separate address for work vs. home) that I hadn't even thought of using Bayesian filtering for anything but spam. Only with spammers can I not trust the simple things! Spammers should [fate far worse than anything I can think of]! -- Luke -- [EMAIL PROTECTED] mailing list
Re: [gentoo-user] Re: [lists] [gentoo-user] Bayesian spam filtering
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 On Thursday 05 June 2003 09:43 pm, Luke Ravitch wrote: > On 2003-06-05 07:15, Anthony Ventimiglia <[EMAIL PROTECTED]> wrote: > > If you mean classification buckets, Bogofilter can be set to classify > mail as spam, not spam (ham), and not sure. That's three. That's what I mean by buckets, Popfile lets you have as many as you want, which I like so I can sort normal mail, work related, mailing lists, spam, etc.. as many categories as I want. -BEGIN PGP SIGNATURE- Version: GnuPG v1.2.2 (GNU/Linux) iQCVAwUBPuANVQqNYTLzAsoIAQKrdAP+Lz6DDm79qJxRGu6iFqse4JYWw/sZB04V qlYxMQs7kMkqFbuk+AmW/cg8Ge8aUAmjZ5xsNpcAL3LxwY+/oETq+z1NyFKCr9/3 lVoWeEpPlEqdUz1TxPnv6T43NLi2oR8NrsTJ+/OC2Iyd53cJOINX777AjlftMs/b HcKbIYSDFOQ= =mQeq -END PGP SIGNATURE- -- [EMAIL PROTECTED] mailing list