[Rspamd-Users] ham/spam classification via forward
christian.mack at uni-konstanz.de
Wed Jun 8 15:15:57 UTC 2022
Am 05.06.22 um 09:11 schrieb lutz.niederer at gmx.net:
> we know that we are able to classify mail as ham or spam by moving it from one imap folder to another and notifying rspamd about that.
> There was the idea of forwarding that mail to specific mailboxes to classify them as ham or spam (and delete after that). This should technically be no problem.
> But in this case the headers are quite new, the sender is another one. We are not sure if the system will learn by that, that the forwarding person is a spammer or hammer. And this would not be the original spammer or hammer.
> What do we get if we implement this? Does it make sense?
We are currently using a forward as attachment for Spam from our
webmailer to special Spam and Ham email addresses.
Only for those two addresses is plugin imap-sieve active.
In that imap-sieve script is an external upload script used
require ["vnd.dovecot.pipe", "copy", "imapsieve", "environment",
pipe :copy "rspamd-learn-spam.sh";
rspamd-learn-spam.sh uses curl to forward those emails to /checkv2.
In order to train those attachments Heinlein did implement another
trigger Header "Rspamd-LearnSpam-Eml: true" in rspamd's /checkv2 for us.
This checks if there is an attached email, if yes then it trains that as
Spam, instead of the complete forward.
Only downside is, that we currently only can set the imap-sieve plugin
for those two accounts, other plugins are stripped.
Kommunikations-, Informations-, Medienzentrum (KIM)
Abteilung IT-Dienste Forschung, Lehre, Infrastruktur
+49 7531 88-4416
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 6007 bytes
Desc: S/MIME Cryptographic Signature
More information about the Users