Status
✅
DRI
Deleted User
Team
📖 Research Enablers
Date
April 24, 2023 → May 10, 2023
Quarter
2023Q2
Subgoal
Goal
This is about evaluating how costly it is to
- Translation to binary hash: create a commR’ evaluated over sha256 (for example) with a proof showing it is the same data than the original commR
- Translation to new curve: create a commR’’ from another proof system tbd passing from sha256 ↔ hash friendly function
Evaluation
Translation to binary hash
Numbers of hash evaluations for 1 Merkle path: 30 (roughly)
Numbers of MT path to open: 3000Using SHA256
Too Costly !
Cost of 2inputs SHA-256 over bls12-381: 40k
Cost for one MT path: 30 * 40k = 1.2M
Total Cost: 3000 * 30 * 40k = 3.6 B
---
USING BN254 Plonk
Cost of 2inputs SHA256: 25k
Cost for on MT path = 25k * 30 = 750 000
Total Cost: 3000 * 750k = 2.25 B
----Benchmark by Celer
Circom: 32M => cost 2-1 sha256 = 32000000/(64*1024/(2*32)) = 31k
Gnark: 45M => 43kUsing Blake2
Too Costly !
Total Cost: 1.8 B Translation to new curve
TBD using new curve !
BLS12-377: cost should be similar
BN254: cost should improve if we can use some plonkish arithmetization.