[Rspamd-Users] Low traffic server. Does it make sense to use spamtrap module?

[G.W. Haywood]

Hi there,

On Thu, 6 Jun 2024, Michael Grimm via Users wrote:

> Hi,
>
> after two month of using and tweaking rspamd settings I am quite confident that 99% of trapped spam is true positive.
>
> I wonder if it makes sense for my two low traffic servers to use the spamtrap madule to train bayes and/or create my own fuzzy storage.
>
> I do run keydb with active active replication, thus combined keys would be available at both servers.
>
> The last month I did receive a total of 5k mails of which 10% became flagged as rejected (quarantined by postfix).
>
> What do you think: does it make sense at all?

These things have a habit of using far more effort for far less reward.

I'd suggest that you quit while you're ahead. :)

My impression of processing with Bayes is that you need to train with
more or less equal quantities of ham and spam.  You don't seem to be
in a position to do that without taking what I'd call liberties with
the statistics.  Perhaps you should be thankful. :)

-- 

73,
Ged.