Does the ZCS do per user anti-spam?

[garithdugmore]

Hello,

I've found a few articles about ZCS' anti-spam and its filters per user but some are old and others unclear.

This bug seems outline exactly what I want to know:

Bug 3870 – per user Spam Assassin score

"I would definitely want the user to be able to mark items they consider "spam"
for themselves, and have that modify their individual score. A Target ad that
Joe thinks is spam may be something Susie wants and does not consider spam. I
think there should be a way for Joe to classify to (train) the system that
this is spam, which would modify his score. Susie would not mark it as spam
and would continue to get the email. Just allowing a user to modify the
thresholds does not give the same flexibility as allowing them to score
individual items that will apply only to them. That allows them to tailor the
spam software and its results to them."

Would anyone be able to tell me if ZCS allows users to teach their own spam bayes db about which mail he/she considers spam/ham?

Currently we use a global anti-spam system where everyone contributes towards spam learning and it ends up being a tug of war. One person thinks an email is spam and another considers it not spam which just confuses spamassassin.

Thanks for any comments in advance

[ewilen]

Right, as the bug indicates, it's not currently implemented.

You might be able to do some per-user customization by editing the amavis and SA configuration files, but I'm sure that's more work than you'd want to do, with a couple exceptions that I'll mention in a moment.

So right now, the only per-user anti-spam is the ability to blacklist/whitelist under Preferences > Mail > Junk Mail Options. I'd suggest encouraging people to use that for edge cases.

What I believe you can do, by modifying config files, is set up some people as spam lovers, who get all mail regardless of score (or maybe bypass scoring, I'm not sure). Do a search on spam lovers in the forums and/or wiki for that.

I think there may also be a way to control who is allowed to contribute to antispam learning. If not, there's probably an RFE on this, but I haven't been able to locate it. The general idea is, suppose you have a doofus who consistently uses the Junk button instead of Delete. You can't fire the person or deny them access to email, so you just make it so that their actions won't cause mail to be forwarded to ham/spam. I suppose a low-tech approach would be to create filter rules in your spam/ham accounts, to just discard emails from certain users.

[garithdugmore]

Thanks for replying Elliot.

So our suspicisions were correct...

What are your thoughts on Dspam?

I see by default Dspam isn't enabled (used the following to enable it -> Using DSPAM for Spam Filtering - Zimbra :: Wiki ) and according to the dspam website -> DSPAM Project Homepage it mentions the following:

"DSPAM is an adaptive filter which means it is capable of learning and adapting to each user's email"

Would you know if this means per user anti-spam?

We've enabled ours and will give feedback on our experiences.

[ewilen]

I haven't yet played around with DSPAM although I've read a few threads on it, and done a little research. This is probably a good place to start reading about DSPAM in a Zimbra-specific context: Enabling DSPAM

I suspect that DSPAM doesn't currently do per-user anti-spam under Zimbra, because it's called by Amavis. This bug seems to confirm my suspicion: Bug 49201 - Potential DSPAM improvements. The initial comment comes from a DSPAM mailing list, from a conversation between a Zimbra developer and one of the chief DSPAM developers.