Really? I figured there would have to be someway to train the filters on the PMG for users other than copying an email and putting it in the blacklist. If you have to have a second filter on the end users server that kinda of defeats the purpose.
Is there a way to make the PMG more aggressive in what it marks as SPAM? We switched from another spam filtering appliance and users are complaining the PMG lets a LOT more spam through.
Best regards,
Eric
On my wishlist for Proxmox to improve PMG is the ability to use it as a full mail server. And/or combining it with a mail archive and secure mail gateway (S/MIME and PGP signing and encryption). Just for the first step, there is not too much to do and many forum users, which don't understand on how to combine with their existing mail server may welcome the possibility e.g. to use a Webmail on PMG to manage the mails, maybe also access the quarantine (as folder) there and maybe be able to access via IMAP as well. Maybe deletions would then not be hard performed just soft performed, so it may be also usable as mail archive. Just by that way, it would be really easy to train spam and ham and improve the system that way. Otherwise their is no easy way to talk back to PMG. Just the ways, which exist are using an extra system to manage extra settings for ham and spam, but finally that makes no sense, rspamd offers a mask, there users or administrator can put their spam and ham in to learn the system, but honestly, it's not user-friendly, neither noob-friendly, quarantine may be used to learn ham by releasing such messages, but I'm no fan of quarantine and also it's missing more important to learn spam. So the only way is to have a spam folder, users can move their spam to and a ham folder, users can move their ham to and fetch that via script to learn spam and ham, other way is to use Sieve and call actions on movements into and out of spam to invoke a script, which pushes this messages to sa-learn on PMG, need to be coded (by yourself) or offer a shared mailbox to put spam and ham inside like a shared Exchange mailbox and learn from there. However, PMG per default does not make sa-learn easy and Proxmox doesn't see the big value in Bayes Database (although it's) but promote to use (better) RBL.
See my Advancing PMG thread about some adjustments with a good set of RBL, additional content checkers for marking more potential spam, but also use sa-learn. My private installation I learned bayes database and have 2-3 spam mails per week maximum, my commercial test installation bayes database isn't learned, so there are more spam mails per week, but form the first complains it got better and better and fetch more and more spam because also of AWL learning, auto-learning, additional RBL and content checks.
