Has anyone integrated Mailman3 with Spamassassin?
What I'd like is a way for emails that I click 'discard' on in Postorius, to be submitted to the spamassassin Bayes classifier to make it more likely similar ones will be rejected in future at SMTP conversation time.
Peter C
Peter Chubb via Mailman-users writes:
Has anyone integrated Mailman3 with Spamassassin?
What I'd like is a way for emails that I click 'discard' on in Postorius, to be submitted to the spamassassin Bayes classifier to make it more likely similar ones will be rejected in future at SMTP conversation time.
I've not heard that anyone has done this. If there's a way to direct email into the classifier, then you can tell Mailman to send discarded emails to some email address, I believe. But I don't know anything about how that classifier is trained. You're likely to get better answers on Spamassassin channels.
To make it one-click would require code changes in both Mailman and Postorius. Not terribly hard as such, but would require some thought as we probably wouldn't want it to be Spamassassin-specific (I think Mark uses SpamBayes on some of his lists, for example, and I suppose rspamd also knows about classifier training).
Steve
-- GNU Mailman consultant (installation, migration, customization) Sirius Open Source https://www.siriusopensource.com/ Software systems consulting in Europe, North America, and Japan
Peter Chubb via Mailman-users writes
What I'd like is a way for emails that I click 'discard' on in Postorius, to be submitted to the spamassassin Bayes classifier to make it more likely similar ones will be rejected in future at SMTP conversation time.
Well anybody who has saved emails classified as spam by spamassassin will have no shortage of sample spam emails. Saving the emails to a Mailman instance will only bring added value if these emails are different from the general spam. All evidence that I have had with spam suggest that spam is sent indiscriminately, i.e. without considering what type of person (or list) is the recipient.
Thus I am not sure this is worth implementing given the other resource constraints that the Mailman developers have.
-- Written by Thomas Krichel http://openlib.org/home/krichel on his 22224th day.
Thomas Krichel writes:
Well anybody who has saved emails classified as spam by spamassassin will have no shortage of sample spam emails. Saving the emails to a Mailman instance will only bring added value if these emails are different from the general spam.
But presumably they are different, since you'd put the spam checker in front of Mailman. If they get to Mailman, then you really want them added to your filter because it would be horrible if they went through and out to the subscribers.
Thus I am not sure this is worth implementing given the other resource constraints that the Mailman developers have.
I don't know either, but it shouldn't be that hard once we know how a couple of trainable filters ingest new spam. That research is the part I'm not interested in doing until at least October. :-)
-- GNU Mailman consultant (installation, migration, customization) Sirius Open Source https://www.siriusopensource.com/ Software systems consulting in Europe, North America, and Japan
"Thomas" == Thomas Krichel <krichel@openlib.org> writes:
Thomas> Peter Chubb via Mailman-users writes
What I'd like is a way for emails that I click 'discard' on in Postorius, to be submitted to the spamassassin Bayes classifier to make it more likely similar ones will be rejected in future at SMTP conversation time.
Thomas> Well anybody who has saved emails classified as spam by Thomas> spamassassin will have no shortage of sample spam Thomas> emails. Saving the emails to a Mailman instance will only Thomas> bring added value if these emails are different from the Thomas> general spam. All evidence that I have had with spam suggest Thomas> that spam is sent indiscriminately, i.e. without considering Thomas> what type of person (or list) is the recipient.
The headers are taken into account by the Bayes classifier. I used to use James Hendrihs solution with Mailman 2: https://www.jamesh.id.au/articles/mailman-spamassassin/
I think I can adapt his mmlearn script ...
Peter C
participants (3)
-
Peter Chubb -
Stephen J. Turnbull -
Thomas Krichel