3 ms·
Worldcoin does the whole thing wrong. What they SHOULD do is to-- 1. Scan biometrics. They should not restrict themselves to just irises, because not everybod
by Hizonner 3y ago
Worldcoin does the whole thing wrong.
What they SHOULD do is to--
1. Scan biometrics. They should not restrict themselves to just irises, because not everybody has irises or eyes. But they should only scan things that you can get surreptitiously without too much work... meaning things that everybody gives away anyway just by walking down the street.
2. Bind the biometrics to keys generated only for that purpose, and never used for anything but attesting to other keys... in zero-knowledge so that you can't associate different sub-keys attested by a given master key.
3. NOT bind the biometrics to names or any other information about the person.
4. Go right ahead and publish the biometrics themselves, together with the associated public master keys. Because again, they are not things anybody should be thinking of as secret... and pretending you can keep them secret is foolish even if you don't put them in a database.
5. Stop centralizing the system that attests to the bindings given in (4); do the work to eliminate trust from that. Or at LEAST don't go out of your way to RELY on a centtral authority.
Once you have that, you have all you need for Sybil resistance, which gives you the global uniqueness they need. Then go build your crypto system on top of it.
- 93po 3y ago1. For now they do collect biometrics with opt-in, because it's valuable data for refining the ability to scan only irises and determine if they're legitimate 2. I can see this being a good idea 3. They don't
- rwiggum 3y agoThis is actually what they're building. They publish ZK proofs over the biometric data instead of publishing the biometric data itself.
- Hizonner 3y agoTheir critical error is that there's a single, central authority that certifies that a person with a given set of biometrics exists. They didn't bite off the hard, but necessary, part where you figure out how walk an attestation graph and come up with a confidence level. And they only do irisis. And they're doing the ZK at the wrong point in the chain. Publishing the actual, raw, unhashed biometrics wouldn't be a problem as long as they didn't tie them to anything but a key. And that future-proofs you.