Hi, Dňa 15. septembra 2022 7:07:23 UTC používateľ Thomas Walter via mailop <mailop@mailop.org> napísal:
>I'd love to use Ham/Spam filters, but in a very diverse environment >(university campus) with users of all languages that's a difficult task to get >right. Sure, in case of big amount of different users, the central SPAM filter can do only basic (obvious) filtering which leads to higher false negative rate (or vice versa, high false positive). IMO that is reason why multiple MUAs has own SPAM filters, to personalize results. But some, of course, is not all, especially not webmails... My user base is small, thus central SPAM filter does good job, of course, with never ending tweaks/improvements (as SPAMs changes). But i tweak it instead of playing online games :-D Anyway, i think, that central SPAM filter still can be useful for you to filter obvious SPAMs, while perhaps not with default settings. >I've seen "fights" between users where one group kept moving emails to their >spam folders while others restored them after the algorithm learned they are >"bad". I am aware of this (and i see it in job), one cannot fully rely on users complains, thus my central SPAM filter is not directly trained by users (i do not use any users contributed RBL too). The SPAM/HAM automation is only for the reporting side. Users mark message as SPAM and i get full (including headers) message copy to decide/act on it. As my false positive rate is very low, it works perfectly. I have idea (not implemented yet) to do server side personal SPAM (bayes) filter, which will be trained automatically (based on move to/from Junk folder) and will eg. negate delivery to Junk folder from central filter. I did some tests in my "lab" and it seems as reasonable, but it still needs more tests and thinking, especially i afraid about performance and resources usage. >If it's only 0,7%, you should be able to press delete a few times more each >day if those land in your inbox. Still better than to miss something important >because it hid in a spam folder. AFAIK, my users are happy with this, but you give me the idea, to make it optional, while still default, behaviour. >Also what about the other 80+% Spam you receive every day? Do you reject those? Sure. I reject these obvious SPAMs on various stages of SMTP dialogue, based on various conditions, mostly scored. At end i use rspamd and i can tell, that after its initial scores/rules tweaks it works very well and its false positive/negative rate is really low. Once message is accepted, it is delivered or processed (eg. DMARC reports are processed by script, not delivered to user). regards -- Slavko https://www.slavino.sk/ _______________________________________________ mailop mailing list mailop@mailop.org https://list.mailop.org/listinfo/mailop
