2 ms·
I'm amazed at how much obvious spam I block from Google. A huge chunk of it is 419 stuff that's trivially caught by SpamAssassin rules that took 5-10 min to wr
by dpifke 4y ago
I'm amazed at how much obvious spam I block from Google. A huge chunk of it is 419 stuff that's trivially caught by SpamAssassin rules that took 5-10 min to write[0].
My guess is that the Gmail team doesn't have any OKRs about spam prevention any more, or else they'd be able to pretty quickly put a stop to the top ~90% of the spam that's currently getting through.
I wish the anti-trust regulators were competent and paying attention. If so, they might be able to do something about the fact that Google—unlike pretty much every other large mail sender in existence—fights really, really hard to avoid taking abuse reports. Gmail is "too big to fail" so we can't just block them outright, like Google would do to us if we were the ones sending a bunch of spam and ignoring abuse reports.
[0] (edited to add) For those who are curious or wish to share the fruits of my labor:
body __DWP_BIZ_PROPOSAL /\b(business|confidential|discrete|mutual|beneficial|profitable|investment|favou?rable)\s+(proposals?|opportunit(y|ies)|friendships?|director(y|ies)|intentions?|introductions?|relationships?|business(es)?|investments?|transactions?)\b/i
body __DWP_DECEASED /\b((deceased|deposed|exiled|late)\s+(former\s+|ex[ -])?(ceo|husband|client|(vice[ -])?president|senator|[mb]illionaire|brother|sister|(great[ -])?((grand-?)?((fa|mo)ther|parent)|uncle|aunt)s?)|widow(er)?|beneficiar(y|ies)|inherit[ea]nce|last will and testament)\b/i
meta DWP_419 __DWP_DECEASED && __DWP_BIZ_PROPOSAL && (FREEMAIL_FROM || FREEMAIL_REPLY)
There are some other rules this works in conjunction with, but stopping the obvious spam is really not rocket science.
- wrycoder 4y agoLOL that was a Procmail flashback from twenty years ago! What happened to Bayesian detection and AI?
- dpifke 4y agoThey work, too. But sometimes simpler solutions work better. SpamAssassin assigns a score (positive = more spammy, negative = more hammy) to each rule. So the above-quoted rules work in conjunction with other rules, the Bayes filter, IP/ASN reputation data, etc. DWP_419 has a really high score in my setup right now because the false positive rate is currently zero, but it could theoretically be canceled out by other signals.