1 Topic by MyKEcz

  • MyKEcz
  • Member
  • Offline
  • Registered: 2014-03-24
  • Posts: 2

Topic: SA-Bayes and separate user tokens

==== Required information ====
- iRedMail version: 0.8.6
- Store mail accounts in which backend (LDAP/MySQL/PGSQL): MySQL
- Linux/BSD distribution name and version: Debian 7.4
- Related log if you're reporting an issue: null
====

Hi guys,
I have question about SA Bayes. I've successfully run IRedMail/FAQ/Store.SpamAssassin.Bayes.In.SQL. Now I want to know how this process works deeper for more users. Every reported spam via Roundcube is saved to DB in form of tokens. Table bayes_vars shows summary of spam, ham and what interests me is user (vmail). Does SA classify spam or ham by user? In this case vmail (global)? What happens if I comment out in /etc/spamassassin/local.cf bayes_sql_override_username vmail? Now every new reported mail will belong to each user and DB table bayes_vars add new user and keep his own settings? Or it just SpamAssasin do not care which user has saved tokens in DB and keep this rules global?
So my main question is how to keep separated tokens by reported users. Because somebody can add to global catalog vmail ham email and it's bad.

I'll be glad for clarification.

Thank you.

Michael

----

Spider Email Archiver: On-Premises, lightweight email archiving software developed by iRedMail team. Supports Amazon S3 compatible storage and custom branding.

2 Reply by ZhangHuangbin

  • ZhangHuangbin
  • ZhangHuangbin
  • iRedMail Developers
  • Offline
  • Registered: 2009-05-06
  • Posts: 31,094

Re: SA-Bayes and separate user tokens

MyKEcz wrote:

So my main question is how to keep separated tokens by reported users. Because somebody can add to global catalog vmail ham email and it's bad.

As you can see in Roundcube plugin config file (roundcubemail/plugins/markasjunk2/config.inc.php):

$rcmail_config['markasjunk2_spam_cmd'] = 'sa-learn --spam --username=vmail %f';

It's a parameter of 'sa-learn' command. So you can simply modify its parameter to achieve your goal.

3 Reply by MyKEcz

  • MyKEcz
  • Member
  • Offline
  • Registered: 2014-03-24
  • Posts: 2

Re: SA-Bayes and separate user tokens

ZhangHuangbin wrote:
MyKEcz wrote:

So my main question is how to keep separated tokens by reported users. Because somebody can add to global catalog vmail ham email and it's bad.

As you can see in Roundcube plugin config file (roundcubemail/plugins/markasjunk2/config.inc.php):

$rcmail_config['markasjunk2_spam_cmd'] = 'sa-learn --spam --username=vmail %f';

It's a parameter of 'sa-learn' command. So you can simply modify its parameter to achieve your goal.

Thank you for fast reply. Okay I'll try to modify this command which will belong in result to specific user for report spam. But my main question was pointed if somebody report spam and I modify this php script to each user, will SA-Bayes arbitrate to specific user's database of tokens or it will all take as "global" spam database and do not care which user is saved id database?

I want to know in result if matter on users tokens or not.

Thanks for explanation. Documentation is weak or I can't find it of SA.

4 Reply by ZhangHuangbin

  • ZhangHuangbin
  • ZhangHuangbin
  • iRedMail Developers
  • Offline
  • Registered: 2009-05-06
  • Posts: 31,094

Re: SA-Bayes and separate user tokens

I think you'd better check SA website to understand how sa-learn works, or do some test on a virtual/testing machine.