[Rspamd-Users] Training Rspamd from Exchange Server

[Carsten Rosenberg]

Am 21.04.20 um 21:59 schrieb Stefan Bauer:
> see my other answer regarding training from yesterday in the list archive.
> 
> for exchange you could have a dedicated mailbox that each user has access
> to and can drop own mails to spam folder.
> 
> then you just grab folder content from remote host via imap/pop and feed
> mails to rspamd.
> 
> Stefan
> 
> Am Freitag, 17. April 2020 schrieb Chris Dos <chris at chrisdos.com>:
> 
>> I looking at creating a mail gateway running Postfix and Rspamd that will
>> be
>> in front of an Exchange 2016 server.  I've been doing research and I have
>> not
>> found any way to train Rspamd with spam/ham.  The Exchange public mail
>> folder
>> method does not seem accessible through imap.  Is anyone able to train
>> Rspamd
>> from an Exchange server?
>>
>>     Chris
>> --
>> Users mailing list
>> Users at lists.rspamd.com
>> https://lists.rspamd.com/mailman/listinfo/users
>>

Hi Chris, Stefan,

@Stefan maybe also about your own topic.

For some Rspamd in front of Exchange projects we have implemented some
functions to process user forwarded mails from Exchange. The user
forwards a mail to a special WL/BL address which is processed by rspamd.
Then an attached message is found, this message will be processed by using

https://rspamd.com/doc/lua/rspamd_task.html#mafbc6
https://rspamd.com/doc/lua/rspamd_task.html#m5ed22

The sender of the attached message will be added to a user based BL/WL.
In general you should also be able to train mails in this way.

But training global bayes with user input is risky. From my experience
users will also train unwanted messages, just because they are lazy.
Maybe think about just to train per-user-bayes in this way.

On the other side it could be a good approach when using fuzzy training
with a low weight. So the kind of spam must been seen several times to
be sanctioned.

--

Carsten