Hello there!
Having a p2p-network to catch mails from honeypot-addresses and training
bayes-nets from that would be good. The From:-address could be
randomized by them, so it could be ignored for training. That would lead
to less problems with learning the addresses of unfunny people. (I‘m
unfunny, so I have an address which is usually blocked by SpamAssassin.
Police used that system. I convinced them not to, because I used that
address to contact them. It was in the Spam-Folder and noone cared.)
That coult be put into Headers as „Spam-Staus: … tests: …“. If I find
that test in the headers i‘d just block it.
Pyzor uses UDP, I guess it doesn‘t train bayes nets, yet. I didn‘t look
into Razor-sites much, but I don‘t think e-mail-clients should check for
spam-staus. Do they treat „From:“ specially?
Regards,
Andreas Wagner
