1

Topic: Report spam from a Maildir ?

==== Required information ====
- iRedMail version:
- Store mail accounts in which backend (LDAP/MySQL/PGSQL):
- Linux/BSD distribution name and version:
- Related log if you're reporting an issue:
==== ==== Required information ====
- iRedMail version: 0.8.1
- Store mail accounts in which backend (LDAP/MySQL/PGSQL): LDAP
- Linux/BSD distribution name and version: CentOS 6
- Related log if you're reporting an issue:
====

Hi Zhang,

I can report spam from a maildir of the user ?

I'm thinking have an account spam@domain.ltd for users to send their notifications.

I review the following url http://wiki.apache.org/spamassassin/ReportingSpam and not work any configuration.

The Regards.

Thanks.

2

Re: Report spam from a Maildir ?

Did you try this tutorial?
http://www.iredmail.org/wiki/index.php? … yes.In.SQL

3

Re: Report spam from a Maildir ?

ZhangHuangbin wrote:

Did you try this tutorial?
http://www.iredmail.org/wiki/index.php? … yes.In.SQL

Hi Zhang

I did it, I do not see any error ... when marked a message as spam, move it to the spam folder and database sa_bayes get different results. but when I send the same email to the same email account appears in my inbox and not in the spam folder, not blocked.

Looks fine:

*********************************************************
Mar 20 12:20:20 xxx roundcube: sa-learn --spam --username=vmail /var/www/roundcubemail-0.7/temp/rcmSALearnuwwFmn
Mar 20 12:20:20 xxx roundcube: array (#012  0 => 'Learned tokens from 1 message(s) (1 message(s) examined)',#012)

mysql> SELECT COUNT(*) FROM bayes_token;
+----------+
| COUNT(*) |
+----------+
|    45758 |
+----------+
1 row in set (0.00 sec)

mysql> SELECT COUNT(*) FROM bayes_token;
+----------+
| COUNT(*) |
+----------+
|    45917 |
+----------+
1 row in set (0.00 sec)

*********************************************************

Regards.

Thanks for advice!

4 (edited by bmackay 2013-03-21 04:42:20)

Re: Report spam from a Maildir ?

Heya ungeme:

Bayesian filters don't work the same as spam rule sets like you would have in MS Outlook.  When you show it something that you say is spam, you are helping to train it to recognize more spam, not set a hard and fast rule that says "never let a message with this content pass".

What would be a better indication of whether your system is working as designed would be to check the HITS value in your log.   Let's say you send a test spam message to your account and it gets a HITS value of 4.5.  Feed that back to your Bayesian filter and tell it the message is spam.  Now send the test message again and see if the value HITS value is higher than 4.5.   If your system is learning, after getting a few similar messages, it will eventually push those spammy messages over your threshold so they get flagged or not not delivered. 

The idea of Bayesian filters is to recognize patterns so that if the spammers tweak a few things in the next email, the filter will hopefully be smart enough to spot it.  Don't expect it to block mail after only seeing one example.  Bayesian works best if there is a lot of traffic to compare.  It's not all that effective on small servers.  If memory serves Spamassassin needs to have at least 200 spam and 200 ham messages in the bayesian database before it does any filtering at all.