Supernova Compressed SNARKs #80
Conversation
There was a problem hiding this comment.
This is a partial review, noting a potential soundness issue. This is proximately due to a potential subtle bug relating to the behavior of zip (truncating results to the length of the shorter of its inputs). It's more broadly due to a failure to explicitly check a condition of the CompressedSNARK that seems critical: specifically, we need to not only check that all provided 'sub-SNARKs' verify, but also that all required sub-SNARKs were provided.
Whether or not the problem I've pointed to is manifested here, let's add a test conclusively demonstrating that the next iteration does not have that issue.
b6e984c to
f1c0040
Compare
|
Since we're moving forward with the v1 |
|
Now that PR #108 is open, I think it's safe to close this PR. |
This PR implements a final compressing SNARK for Supernova NIVC proofs. It also adds a test for the compressed SNARK.
This version 0 of compression simply generates a SNARK proof for each primary running instance, and the secondary instance. The verifier then checks the validity of each of the SNARKs.