Nostr Web Client

Spam prevention should work by discarding events from users that are not connected to your social network like normal users are. For example, swarms of users that are connected by only 1 or less follows from legit users.

Iris already discards messages from completely unconnected users, but need to add shared block list support and better detection of sybil swarms.

This should work especially for trending algorithms, where reactions from socially connected users are more valuable anyway, and missing a like doesn't matter so much.

Currently Iris trending list comes from nostr.band, but I want to do it locally at some point. That's the only way you know what you get.

waxwing 2y ago

Anti sybil based on pattern recognition, or reputation, will always ultimately be gameable.

Real costs (proof of work or micropayments, a la hashcash) are the right direction imho.

Reply to this note

Please Login to reply.

Discussion

j 2y ago

pow

nostr:nevent1qqs0sqvsfrumtj4fneff0g7mrnp4cfa5j9sq23lp3vscnq92lees6eqpz3mhxue69uhhyetvv9ujuerpd46hxtnfdupzqe6msnl8tcsk4w28capcaegeefmh2dmdmuza4ha6vfut6qfwr4egqvzqqqqqqyl0q48l

sommerfeld 2y ago

Social connections is PoW. It requires energy to produce content and get real people to follow you.

waxwing 2y ago

But it's not easily (or even, at all perhaps) verifiable.

A hashcash hash is verifiable completely independent of context, and it's almost ridiculously easy/cheap to verify.

waxwing 2y ago

It's also very quantifiable (at scale, where the law of large numbers applies).

Prosocialise Foundation 2y ago

As if people didn't already pay billions to access and influence others...