rSpamD-Training via GUI

Feature
, ,
1

Following the recommendation here, I would like to make a suggestion.

In particular, after a new system has been put into operation or after the rSpamD filter has been reset, a long period of manual training may be required to teach the filter.

During this time, users are confronted with an increased amount of spam and there is a high risk that the filter will be trained incorrectly because users do not consistently move spam emails to the spam folder.

Relying on or relying on correct user behavior usually leads to failure.

That’s why I’ve been using the following procedure for years under NS7:
https://wiki.nethserver.org/doku.php?id=userguide:manual_training_of_the_bayes_filter_rspamd

I haven’t been able to adapt it for NS8 yet, and I’d rather suggest GUI support for it anyway.

To do this, the adminsitrators would only have to download the required samples from
http://untroubled.org.

The GUI should offer an upload option and the 7zip files should be unzipped automatically in the background and the training should take place.

At the same time, a reset button should be provided for the filter to restart this process. Experience has shown that filters deteriorate over time because users classify unreliable and this misconduct corrupts the filter system-wide.

Alternatively, it would be desirable if each user / mailbox had their own filter, so that not everyone would suffer as a result.

In that case, the above-mentioned function would have to be located directly at the mailboxes or email accounts.

5 Likes
2

A more straightforward solution would be to automatically learn 400+ hams from a user-selected INBOX or even an automatically chosen set of Inboxes.

For the spams I agree that it is often necessary to rely on a set of external samples but still the system should be able to learn form users’s spam folders to get a more accurate dataset (especially for foreign language users).

1 Like
3

I agree with you, a GUI-based solution would also be desirable for that.

But it can only be used under two conditions:

  1. you have a mailbox with +400 mails. This is not the case with new servers/organizations.
  2. you actually have an e-mail account with pure HAM mails, otherwise you can save yourself the trouble. For this you would have to validate every e-mail in advance.

I can’t even vouch for all my e-mails - who is immune from careless mistakes.

Yes, but that can take quite a while and requires minimally competent users who are willing to cooperate. Just thinking about my environment, I wouldn’t want to rely on it.

And in a company, users don’t care at all. After all, IT is there for the user and not the other way around.

Deleting is much easier and you don’t have to think about it every time.