Hacker News new | ask | show | jobs
by x4e 1829 days ago
Do you have any reading material about how you achieved this?

I can't really see how zero-knowledge proofs could solve this. There is no cryptographic way to prove that software executing on a clients machine triggered a notification. Especially on Linux where an open source notification manager could be modified to reject it.
2 comments
freebuju 1829 days ago
Assuming you have gone through this [0] and it did(n't) click for you.

I'm equally not so convinced on this anonymous ad system they claim to have built. The browser claims to generate an adID based on your history but encrypt this info to the advertiser. Maybe someone who has actually interacted with the ad platform can provide more insight on what information is exposed.

Zero-knowledge advertising sounds practically like an oxymoron to me, but hey they claim to have made it work.

[0] https://brave.com/themis/

link
jonathansampson 1828 days ago
Certainly! Check out the resource detailing our Ad Confirmation process at https://github.com/brave/brave-browser/wiki/Security-and-pri... (it's a little old, but should be helpful). We leverage the Privacy Pass approach too, so reading https://www.petsymposium.org/2018/files/papers/issue3/popets... will also help understand our process. I hope this helps!
link
x4e 1827 days ago
Perhaps I am misunderstanding what you sent, but isn't this just a way for the user to report that they viewed an add, not prove that they viewed it?
link
jonathansampson 1826 days ago
The cryptographic proofs are baked-into the confirmation and reporting process.

A sufficiently-capable attacker could conceivably trick the browser into thinking a native OS ad-notification was displayed, we do rely on the OS to inform us at this point (though preview versions of Brave do not have this dependency), but we have considered this as well.

The main threat here would be an attacker who attempts to automate the confirmation process, and potentially duplicate it across various VMs or OS instances. Fortunately, we've considered this too. For reasons I hope are obvious, I can't go into greater detail here.

link