In POSTSCREEN i use 12 blocklists and whitelists. each is given a
particular score based on a custom ML algorithm. The scores get adjusted
everyday based on the performance of the RBLs. I don’t reject based on
SPF, DMARC, or DKIM.
However i do have spam detection powered by RSPAMD, which takes into
account SPF,DKIM,DMARC and host of other stuff. right now experimenting
with LLMs as tool to detect SPAM apart from the standard baye’s.
On 2024-07-30 11:52, Jaroslaw Rafa via Postfix-users wrote:
Dnia 30.07.2024 o godz. 12:38:15 Matus UHLAR - fantomas via
Postfix-users pisze:
>I filter messages only based on RBLs, manual blocklists and content
>filtering (SA + many custom rules). And as for the latter, the messages are
>sent to spam folder, never rejected. Rejections are based only on first two.
Funny, since multiple people in the past recommended rejecting on
spamminess, not on the results of single DNSBL listing.
I use only two DNSBLs that - at least for me - seem to give almost no
false
positives - bl.spamcop.net and zen.spamhaus.org. In the past I used
three -
instead of zen.spamhaus.org I was using dul.dnsbl.sorbs.net and
sbl-xbl.spamhaus.org. But because sorbs.net went down, and
zen.spamhaus.org
seems to effectively combine these two, I changed it.
Of course I always have the option to manually override DNSBL listing
in my
manual access list, but I don't remember when I last had to use it.
_______________________________________________
Postfix-users mailing list -- postfix-users@postfix.org
To unsubscribe send an email to postfix-users-le...@postfix.org
