3 ms·
It really depends on the implementation around it, how a user conducts themselves, and what data you can buy. While there is zero knowledge inside the proof, it
by Tadpole9181 1y ago
It really depends on the implementation around it, how a user conducts themselves, and what data you can buy. While there is zero knowledge inside the proof, its use creates a side channel that reveals information.
For instance: The relying party server needs to call the auth server on novel users. Thats a new, unavoidable indicator!
How large are token batches and how long do they last? Will the implementation force them to wait a time period between redemption and use?
A bad implementation means the user IP will talk to the A server, then it will contact the RP server, who will contact the A server. Because this happens once per connection (or 60 minutes in this bill) and takes maybe a few hundred milliseconds. there's not going to be a huge number of candidates to have to sort through. And that's just the handshake.
- bawolff 1y ago> For instance: The relying party server needs to call the auth server on novel users. Thats a new, unavoidable indicator! Not really. There is no requirement here for an auth server to neccesarily even exist. That said, your broader point is correct, that the details matter a tremendous amount.
- Tadpole9181 1y agoOh, that's my bad, I re-read the privacy pass protocol to brush up and it does use signing without requiring the RP to necessarily make another call to the original approver server. I also see there's been work on hidden witness ZKP, so the RP may not even know who approved a given token. Very cool! Always happy to be proven wrong with cool tech!