On Tue, Oct 15, 2019 at 4:08 AM Jaroslaw Rafa <[email protected]> wrote:
> Dnia 14.10.2019 o godz. 13:23:30 Brandon Long via mailop pisze: > > > > Modern spam filters are a combination of good and bad signals, but if you > > have no good signal... then we only have the bad ones. > > Well, I could see at least few potential "good" signals - I wonder if you > use these, and if yes, why don't they overcome the vague indication of "bad > neighborhood". > > 1) the sending server consistently uses the same sender e-mail addresses > (and most of all one address) - these aren't multiple random addresses like > spammers usually do. You can see a strong, clear correlation and > consistency between the sending IP and the sender e-mail addresses used. > eh, not a great signal. Lots of spam comes from the same sender email address, especially if you consider more "grey" spam types (ie, marketing mail) 2) Gmail users are actually engaging in e-mail exchanges with the sender, > they do reply and emails are sent back and forth instead of just deleting > or > ignoring the message > Yes, and that is a signal that we do use. It's also typically why you are less likely to go to the spam label when going back with the same people. OTOH, you'd be surprised the lengths spammers will go to to try and games this signal up front before their campaign. 3) I'm subscribed to several mailing lists hosted on Google Groups, I write > quite a lot of messages to these groups (using the same sending IP and > e-mail address), these are accepted and distributed and don't go to users' > Spam folders. Spam washing through Google Groups is it's own problem. We could probably use a signal there using ARC to pass through the original authentication, but using the ARC signals is still a work in progress. Brandon
_______________________________________________ mailop mailing list [email protected] https://chilli.nosignal.org/cgi-bin/mailman/listinfo/mailop
