Vitalik Buterin on Warpcast

Content pfp

0 reply

0 recast

0 reaction

Vitalik Buterin pfp

Vitalik Buterin

When debugging a ZKP verifier, make sure to keep checking not just that the proof verifies on valid inputs, but also that the proof verification returns an error on invalid inputs (or with nonsensical tweaks to the verification mechanism).

77 replies

911 recasts

3799 reactions

Emmanuel Awosika pfp

Emmanuel Awosika

I guess this is how you catch unconstrained circuit bugs? I can imagine there's some proof circuit that checks that the square of a number is 9, but doesn’t check if it should be 3 or -3. What you're suggesting here should catch the bug if we only want it to be 3, no?

1 reply

0 recast

15 reactions

TongueScraperLagrange pfp

TongueScraperLagrange

@eulerlagrange.eth

Hey Vitalik would you like to do an AMA on here sometime?

4 replies

42 recasts

297 reactions

0xqeew pfp

In zk proofs, there are (2) actors : prover and a verifier. The prover convince the verifier without revealing any information. The verifier verify that the prover is telling the truth. A ZKP verifier must reject both invalid proofs and those with tampered verification.

0 reply

4 recasts

35 reactions

jenny.degen 🎩🟣 pfp

jenny.degen 🎩🟣

Great point, @vitalik.eth! Ensuring that a ZKP verifier properly rejects invalid inputs is crucial for maintaining security and integrity. It's not just about accepting the right proofs, but also about robustly identifying the wrong ones. 🔍🛡️

5 replies

6 recasts

24 reactions

Majid pfp

0 reply

1 recast

13 reactions

non 🐹 pfp

Yes, sir, I will🫡

0 reply

1 recast

13 reactions

WOO🎩 pfp

Yes sir thx Legend 🙇‍♂️

0 reply

1 recast

11 reactions

Emp pfp

@empathicalchemy

I think the core of this, is crucial in so many things. Look for the truths in things, the provable, but also the untruths, the unproven. 10 $DEGEN

0 reply

1 recast

5 reactions

pgpg pfp

I mean. That's just general testing advice sir.

0 reply

0 recast

4 reactions

Giancarlo🎩 pfp

@giancarlodc.eth

Good unit tests should be able to take care of that or am I missing something? Still, ensuring to cover all these use cases is what I am calling “good” unit tests.

0 reply

0 recast

4 reactions

sirlupinwatson.degen pfp

sirlupinwatson.degen

@sirlupinwatson.eth

So in theory could you for example instead of validating it, you would overflow not bytes but let's say for: 32-bits max -2³¹...2³¹ 64 bits max -2^63...2^63 double precision floating point 1.8 x 10^308 Could you overflow (not sure if this is the right term) with a constructor and use yobibyte or bs, from there you could technically infinitely drain a process?

1 reply

0 recast

0 reaction

ap9671 pfp

Insightful as always!

0 reply

0 recast

0 reaction

elysiagh pfp

0 reply

0 recast

0 reaction

Phanthuc pfp

0 reply

0 recast

0 reaction

barnabyye pfp

0 reply

0 recast

0 reaction

Ben Green pfp

I feel like this message is a direct response to snarkjs and its lack of testing. This pr is in the new 0.7.4 and it has no test that fails before the patch. https://github.com/iden3/snarkjs/pull/480

1 reply

0 recast

0 reaction

Anatole pfp

0 reply

0 recast

0 reaction

Minhng83 pfp

0 reply

0 recast

0 reaction

Puppy pfp

0 reply

0 recast

0 reaction