@varx @rodolforg no, plus horribly inefficient. But also, most of this spam wasn't to "mentioned only" audience, but was actually publicly posted, just at mentioning or replying to an account.
Post
@thisismissem @varx checking display name (including unicode codepoint aliases and the invisible ones) and profile photo would help a lot, I believe
@rodolforg @varx and how do you "check" the profile photo? Are we going to run them through PDQ or PhotoDNA?
Sure, we could keyword match in display names for a list of words, but then you'd flag "Ivory for Mastodon" for example, if "Mastodon" was on the keywords, even though that's not a spam attempt.
@rodolforg @thisismissem probably, im assuming pairing that with any patterns they can find in the PII is why @thisismissem is asking for that data. Categorizing spam by content though is such a cat-and-mouse game.
I just discovered the idea of proof-of-work stamps to help curtail spam in #reticulum so Im in the "everything is a nail" phase of hyper-focus. To her point though, you *maybe* could require that for "DMs" but it would be totally unreasonable for all public posts.
