It'd be neat if we had software relay operators could run that would crawl all npubs in their database (it could behave just like any other client requesting notes) and assign something like a "trust score" to each.

It could use some web of trust heuristics and spam reports, possibly seeded by a few well established users. Relay operators could then use those scores for spam filtering or cleanup jobs deleting high probability spam.

The main benefit is this would be a generic tool any admin can run regardless of how their server is implemented.

Does this exist? #asknostr

Reply to this note

Please Login to reply.

Discussion

Moderation tools for relay operators are hard to come by. That said, nostr:npub1jlrs53pkdfjnts29kveljul2sm0actt6n8dxrrzqcersttvcuv3qdjynqn recently released Landlubber, which is a web gui for relay moderation, and I think Coracle was the first client to implement WoT filtering, so I wouldn't be surprised to see some kind of WoT-based moderation tools added to Landlubber. Though I think nostr:npub1kpt95rv4q3mcz8e4lamwtxq7men6jprf49l7asfac9lnv2gda0lqdknhmz's pagerank might be a more appropriate starting point for what you are talking about.

it's me

nostr:nprofile1qqstq4j6pk2sgaupru6l7ah9nq0dueafq356jllwcy7uzlek9yx7hlspz4mhxue69uhhyetvv9ujuerpd46hxtnfduhsz9mhwden5te0wfjkccte9ec8y6tdv9kzumn9wshsz9mhwden5te0wfjkccte9ehx7um5wghxyctwvshs4kprv3 would you be interested in providing an http endpoint that can query the entire social graph as related to an 'observer'? IE a list of keys that have a certain rank. It's usually a very large list (100,000 pubkeys) is why DVM cannot fit it into an event... Or do you have some kind of DVM pagination that could do it with many DVM requests?

sorry I missed this message.

At the moment we don't offer this functionality, but we have RankProfiles, where you can request the rank of up to 1000 pubkeys in a single call.

https://vertexlab.io/docs/services/rank-profiles/

A similar idea would be to return a bitset containing all pubkeys whose rank is higher than a threshold according to an observer/source (for example using sort=personalizedPagerank and source=your npub).

What is the usecase that u have in mind?

My usecase is to prevent spam on relays. The relay would query for a large set of pubkeys and use it for whitelisting. I've been finding that any keys above a (usually low score threshold) helps deter spam and scams. 1000 isn't enough.. I've been using lists of ~100,000.

Are you using something like? https://github.com/mmalmi/nostr-social-graph

Using https://github.com/Pretty-Good-Freedom-Tech/brainstorm. It's a bit better than the follows of follows methods.