Topic: BayesFiltering for SpamAssasin

[d_gerst]

Hello,

That correct, folder will be automatically created. My script will check if folder exists. I would apologize for not thinking about.

Danke schön für den netten Hinweis, mercyh. Hab ich garnicht im ersten Moment dran gedacht.

But did someone see logfiles that produced from spamd on SME7. I still missed the output how long it took to scan mails with bayes? On SME6 productivity it will be generated.

Best regards,
Daniel

[d_gerst]

Hello,

There's a little bug in the crontab file it should be 30 0 * * * *.

Please Reinstall the script or modify the file:
1.) /etc/cron.d/sa-bayes_learning
2.) service crond restart

Sorry!

Best regards,
Daniel

[cool34000]

You were right azche24, I went to /etc/mail/spamassassin and launched manually the script bayes_filter.sh

Folders were created !!! As it's a test server, it was not online yesterday at 0:30, so folders couldn't be created anyway !!! Bad luck !!!
All is OK in the webmail now !

Thanks to you d_gerst, by default outlook don't show all dirs !
I went to my SME IMAP inbox in Outlook, there is an option in IMAP FOLDERS : a box that need to be unchecked. It's something about "only showing subscriptionned folders". I still don't know what are those types of folders, but yours aren't ! Something like newsletters maybe ? Whatever, this is not a topic about IMAP !

Finally, it works ! So thanks to everyone for helping and sharing !
... Just hope one day I can help someone too

[cool34000]

I want to rename folders because their names aren't very explicit in french...

First of all, i want to be sure that i understand well...
junkmail is spam folder
junkmailmissed is where i put my false negative spam
hamlearning is where i put false positive spam
Is that right ?

Anyway, I edited bayes_filter.sh and found the these lines :

Code: [Select]

ham='/home/e-smith/files/users/'$user'/Maildir/.hamlearning'

Code: [Select]

spam='/home/e-smith/files/users/'$user'/Maildir/.junkmail'

Code: [Select]

missed_spam='/home/e-smith/files/users/'$user'/Maildir/.junkmailmissed'

I want the folders to be named respectivly :
.SPAM faux positif
.Courrier indésirable
.SPAM faux négatif

Really easy you gonna tell me... But how can I put spaces in folder names in the script ??? It considers everything else after my space as an arguments... What's the good syntax ?
I'm not trained with Linux syntax and it does not looks like m$ !!!

And is that possible ? I mean for example, if I use the "auto-sort junkmail to junkmail folder" in sme's server-manager, will it works ?

[Smitro]

I was just about to try this out, then I thought... wouldn't it be good if we could compact this down a little. Make it easier for people with more than the average amount of folders.

Is it worthwile changing the layout of the folders mentioned to something like:

Code: [Select]


Junkmail ->
            - Junkmail
            - Spam Learning
            - Ham Learning

So one folder with 3 subfolders.  Or junk mail could be in the top folder and there could be 2 sub folders.

Just a suggestion... thought it might clean peoples folder up a little.

I don't get enough junk that I need to use these folders every time I login.

What do you think?

[cool34000]

That would be great !

[Smitro]

d_gerst, your thoughts?

or, could you let us know how it's done.

Thanks.

[azche24]

Hi,

is there a way to delete the learned spam or ham in the folders created and used by this script?

This does not work here  

And the "LearnAsSpam" script another user posted here could do that.

[cool34000]

There is a cron to delete junkmail, don't know if it's activated by default, but maybe you could look there...

[azche24]

Hi,

There is a cron to delete junkmail

no:

Code: [Select]

/etc/cron.d/purge_junkmail

and it´s related script only cleans up the ../user/../junkmail folders.

I have to clean up my "junkmailmissed" and "hamlearning" folders. purge_junkmail does not do that.

[XAPBob]

Just a thought - it might be nice to move the files to *.bup instead of deleting them - and to kick the relevant script:

I have checked the firewall logs & smoothwall is definitely blocking the connect.

from the install script.

And I picked up:

Code: [Select]

Argument spam
netset: cannot include 192.168.1.25/32 as it has already been included
ERROR: configuration specifies 'use_bayes 0', sa-learn disabled

Learn missed spam...
No missed spam files available.

------------------------------------------------------------------------------

Show statistics...

ERROR: Bayes dump returned an error, please re-run with -D for more information
------------------------------------------------------------------------------


Which is peculiar - the script tries to set UseBayes, the second script then complains about use_bayes

[robwellesley]

So here's an idea,

Is it possible then to create a user called learnspam and have other users forward spam to that user, and have the spamLearner check only that mailbox?

This would work better for pop3 users.

Same for Ham.

cron to delete mail nightly.

Rob

[brianr]

sounds possible, however when this has been discussed before, it has been pointed out that SA might "learn" than spam is email which is forwarded, i guess the underlying issue is whether the headers from the original email are preserved and also how the body is forwarded.  I understand that some (proprietary) email clients are a bit lax in this area!!

[robwellesley]

Yes,
I've since seen this...

http://spamassassin.apache.org/full/3.0.x/dist/doc/sa-learn.html#effective_training

...Point 4 - which is much better as folk are more likely to be motivated to forward incorrectly identified HAM as it assists them.

In my experience Most users are far to busy to engage in sorting SPAM/HAM into separate folders

[hardijs]

I have a question - does the script "manage" the junkmailmissed (etc) messages - like erasing them or is it the responsibility of the one who put it in there?
I am gettin an error in the daily report to the admin:

Code: [Select]


Clean up spam messages from learning folders
/etc/mail/spamassassin/bayes_filter.sh: line 152: /bin/rm: Argument list too long


Is this filtering userlevel (ie each user will have its own spam training) or is it systemwide - ie one user trains - all users benefit?
So  the system to recognise messages from x to be spam whereas some other user receives and is expectin that as x  as a news source