161–170 of 286 posts
I have exactly the opposite problem: tons of legitimate email is getting flagged as Spam by GMail. It's now happening regularly with emails from people in my contacts with whom I regularly exchange messages . Mind-boggling. I know spam filtering is a hard problem, but these are just obvious misses.
I feel like it used to be near perfect. Something must’ve broken the models.
Earlier quoted context omitted.
Who is to say that it isn't working as intended in building a higher wall around their ad garden? There should be a way for them to accept mail from self hosted users, or even provide some sort of testing tools (perhaps validated by credit card some type of bar to increase accountability? At least an RBL or something) for those who do, but nope.
I am beginning to think that the simplest explanation is the most likely one. Google is just bad at stuff... If this was all part of some clever conspiracy then it would be, well, more clever...
Another kind of spam mail like this is they use google docs or sharepoint to add a huge CC list.
However, defending them is hard and I think I can share my knowledge on topic of spam filtering(given I run an email forwarding service (https://mailwip.com) and have to deal with spam a lot)
When an email come from gmail/hotmail(any popular free email service) itself, it's harder to detect spam, especially if the email is in a non English language.
It has a few way to flag spams:
- Look at the IP address: - Look at the structure of emails: follow best practice, such as html/text plain part, has right mime encoding etc - Look at header of emails: no weird header, no "bad" ip in received chains - Look at attachment file type, virus scan those attachment - Finally, tokenize content of the email to find similar email that are flagged as spam
When the email come from their own IP, send out by gmail themselves, email format looks good, DKIM/SPF all pass and this is the first kind of email then the only way to flag spam is by analyze content. And if the email is in non English language, it's harder to analyze. Especially if not enough people flagged it as spam to train the naieve bayes tokenizer then we're out of luck here. The long CC list looks like a legitimate indicator for spam, but librarian/scool has a tradition of sending out a huge CC of entire class/department, sometime even BCC which make the email looks very suspicious (undisclose recipient) yet they are legitimate email so the CC alone cannot easily be used for spam indicator.
Yet, at the same time, legitimate emails form your own server get flagged because low reputation or a history of previous owner send spam...
Earlier quoted context omitted.
You live in a different universe than I do. Fascinating. In my world, spam was a true problem that was killing email. Truly killing it. You would have to create a new address every 6 months to stay ahead of the spam and have a functioning inbox. Very different world. Gmail fixed that. And for me, there have been issues from time to time and in fact i am getting this very spam myself, but gmail permanently stopped the…
>Gmail fixed that. Hardly. They didn't invent the tech. And they didn't perfect it. They haven't even tried. The evidence is in my Spam folder. I mean they mark messages they themselves originate as dangerous spam. "Dangerous" being their word, not mine. Screenshot from within the last hour: https://imgur.com/IjQUGaY I was in that earlier world too. They papered the problem over, a little. But no, they didn't fix it.