[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Notice: in file [ROOT]/includes/session.php on line 2208: Array to string conversion
[phpBB Debug] PHP Warning: in file [ROOT]/includes/functions.php on line 4688: Cannot modify header information - headers already sent by (output started at [ROOT]/includes/functions.php:3823)
[phpBB Debug] PHP Warning: in file [ROOT]/includes/functions.php on line 4690: Cannot modify header information - headers already sent by (output started at [ROOT]/includes/functions.php:3823)
[phpBB Debug] PHP Warning: in file [ROOT]/includes/functions.php on line 4691: Cannot modify header information - headers already sent by (output started at [ROOT]/includes/functions.php:3823)
[phpBB Debug] PHP Warning: in file [ROOT]/includes/functions.php on line 4692: Cannot modify header information - headers already sent by (output started at [ROOT]/includes/functions.php:3823)
Poco Forums • View topic - Junk filter running wild
Page 1 of 1

Junk filter running wild

PostPosted: Wed Oct 08, 2008 5:29 pm
by Victor Dewsbery
Somehow, the junk filter in Barca is running wild. This morning, when I downloaded my mails, about half of the regular mails were sorted into the junk filter. The one or two junk mails that arrived were left in the IN box.

Most of the false positives in the junk filter were from mailing lists that I subscribe to (mainly Yahoogroups and JiscMail). But other mails from the exact same groups landed in the IN box. But the junk filter also sometimes catches false positives which are important (and usually urgent) mails from my clients. So the junk filter seems to be more trouble than it is worth.

Has anyone else observed the same problems?

~ Moved from Barca ... Eric ~

A closed book

PostPosted: Wed Oct 08, 2008 7:54 pm
by Victor Dewsbery
I saw this forum ("Junk Mail Filtering ...") after I had posted, and I am happy to note that a moderator has moved my post here.

I read through a couple of posts in this forum. As far as I understand, the filter is unable to allow individual domains unless I cripple all "intelligent" junk checking procedures. The messages in question spoke about "Bayesian" and "non-Bayesian" filters (or something like that), and I didn't understand a word of what they were saying. (':roll:')

I can confirm, however, that allowing domains (for my mailing lists) and the function "Classify as good" are about as effective as trying to dam a river with a pencil. (':lol:')

Can anyone recommend a decent external junk filter?
My ISP has a useful filter service that holds back most stuff, and until recently I also used Cloudmark spam filter (as a secondary defence against the stuff the ISP filter misses), but that only works for Outlook.

Re: A closed book

PostPosted: Wed Oct 08, 2008 8:49 pm
by Eric
Victor Dewsbery wrote: The messages in question spoke about "Bayesian" and "non-Bayesian" filters (or something like that), and I didn't understand a word of what they were saying. (':roll:')
There's a Bayesian filter in Barca which you need to check in order to work (See Junk Mail filtering in Barca). It needs at least 1,000 good and bad words to intervene. :)
Further more there are some filters active which don't use Bayesian filtering. When you open your filters it should be the last in the list. If not, move it down.
I can confirm, however, that allowing domains (for my mailing lists) and the function "Classify as good" are about as effective as trying to dam a river with a pencil. (':lol:')
Never used that function, so I'm not aware if it works or not.
For mailing lists I do put them into my Contacts (> Subscriptions), so they won't be regarded as spam. :P
Can anyone recommend a decent external junk filter?
See Spam filtering app that works well with Pocomail?.
Spamihilator, K9 and others mentioned in that thread are free and do work quite well. 8)
My ISP has a useful filter service that holds back most stuff, and until recently I also used Cloudmark spam filter (as a secondary defence against the stuff the ISP filter misses), but that only works for Outlook.
My ISP's filter deletes 'spam' automatically, so I de-activated that filter. Lost some pretty valuable information about my purchases & other stuff with that filter in the past. :x

Re: A closed book

PostPosted: Wed Oct 08, 2008 11:28 pm
by Victor Dewsbery
Eric wrote:There's a Bayesian filter in Barca which you need to check in order to work (See Junk Mail filtering in Barca). It needs at least 1,000 good and bad words to intervene. :)

':?:' ':?:' ':?:' ':?:' ':?:'
Is there no hope for a Barca newbie? Where am I to conjure a list of good and bad words from (and in which language)?

Further more there are some filters active which don't use Bayesian filtering. When you open your filters it should be the last in the list. If not, move it down.

List? What list? I have tried 2 menu items (Tools>Filters and Tools>Junk Mail Filters) and all I see is tick boxes for the Bayesian filter. No other filter anywhere to be found.
What other corners of the program should I look in?

Can anyone recommend a decent external junk filter?
See Spam filtering app that works well with Pocomail?.
Spamihilator, K9 and others mentioned in that thread are free and do work quite well. 8)

Perhaps I will explore them further, although the discussion in that thread seems to suggest that they are similarly geekish.

My ISP's filter deletes 'spam' automatically, so I de-activated that filter. Lost some pretty valuable information about my purchases & other stuff with that filter in the past. :x

Mine puts them in a "Spambox" on the server, and I go there and delete them (and retrieve any false positives) every few days.

Re: A closed book

PostPosted: Wed Oct 08, 2008 11:43 pm
by Eric
Victor Dewsbery wrote:':?:' ':?:' ':?:' ':?:' ':?:'
Is there no hope for a Barca newbie? Where am I to conjure a list of good and bad words from (and in which language)?
The Bayesian filter will add those words automatically or if you classify mails good or bad.

Image

List? What list? I have tried 2 menu items (Tools>Filters and Tools>Junk Mail Filters) and all I see is tick boxes for the Bayesian filter. No other filter anywhere to be found.
What other corners of the program should I look in?
Open your Filters (incoming), not the Junk Mail Filter (= Bayesian)
Perhaps I will explore them further, although the discussion in that thread seems to suggest that they are similarly geekish.
As with all Bayesian filters they need to learn, so it will take some time, although not that long. :)
Mine puts them in a "Spambox" on the server, and I go there and delete them (and retrieve any false positives) every few days.
Ah, that's a better approach. Mine were automatically deleted. Spam or no spam, they were gone. :shock:

Re: A closed book

PostPosted: Thu Oct 09, 2008 12:16 am
by Victor Dewsbery
Eric wrote:
Victor Dewsbery wrote:':?:' ':?:' ':?:' ':?:' ':?:'
Is there no hope for a Barca newbie? Where am I to conjure a list of good and bad words from (and in which language)?
The Bayesian filter will add those words automatically or if you classify mails good or bad.

So I have to enter them on the fly and wait until I pass the threshold. What happens until then? And how does the filter learn the words if I only feed it mail messages wholesale, not word by word?

Image

Yes and yes.

List? What list? I have tried 2 menu items (Tools>Filters and Tools>Junk Mail Filters) and all I see is tick boxes for the Bayesian filter. No other filter anywhere to be found.
What other corners of the program should I look in?
Open your Filters (incoming), not the Junk Mail Filter (= Bayesian)


Been there, done that, bought the postcard, and there is still no other filter given in the list.

By the way, since I upped the settings of the "learning Bayesian filters" (junk threshold 0.99, good mail bias 3, junk score 100, good score -100) and reactivated junk filtering, I don't think there has been a false positive (although there have not yet been any junk messages either). So these controls seem to do something (the default settings gave me way over 50% false positives). I'll see if there is any change in the results when I download the overnight messages in the morning.

P.S.:
Oops, :o :o :o looking at Tools>Junk Mail Filters>Status, it seems that I have gathered 2,693 junk words and 10,626 good words. But I haven't a clue what that means, where they come from or what the filter does with them. The status box also claims that there are 18 filters operating, but they are not listed anywhere (so my comments on the missing list of filters above still apply).

Re: A closed book

PostPosted: Thu Oct 09, 2008 1:33 am
by Eric
Victor Dewsbery wrote:So I have to enter them on the fly and wait until I pass the threshold. What happens until then? And how does the filter learn the words if I only feed it mail messages wholesale, not word by word?
Nope, when you've activated the Bayesian filter it will fill in those words when receiving Junk, or when you classify a message as good. :)

Been there, done that, bought the postcard, and there is still no other filter given in the list.
Did you have a look herer?

Image

You can simply deactivate it by unchecking that box or leave it as it was. This filter should be at the end of all your filters.

By the way, since I upped the settings of the "learning Bayesian filters" (junk threshold 0.99, good mail bias 3, junk score 100, good score -100) and reactivated junk filtering, I don't think there has been a false positive (although there have not yet been any junk messages either). So these controls seem to do something (the default settings gave me way over 50% false positives). I'll see if there is any change in the results when I download the overnight messages in the morning.
I agree, you need to play with the settings. If too low, to many spam will pass through. It's not yet ideal, so perhaps that will change.
On the other hand spam techniques do change over time, so it's not easy to get them all. :?
The status box also claims that there are 18 filters operating, but they are not listed anywhere (so my comments on the missing list of filters above still apply).
That's the filter as in the screenshot. :)