|
/*
|
|
Sonobe's Nova + CycleFold decider verifier.
|
|
Joint effort by 0xPARC & PSE.
|
|
|
|
More details at https://github.com/privacy-scaling-explorations/sonobe
|
|
Usage and design documentation at https://privacy-scaling-explorations.github.io/sonobe-docs/
|
|
|
|
Uses the https://github.com/iden3/snarkjs/blob/master/templates/verifier_groth16.sol.ejs
|
|
Groth16 verifier implementation and a KZG10 Solidity template adapted from
|
|
https://github.com/weijiekoh/libkzg.
|
|
Additionally we implement the NovaDecider contract, which combines the
|
|
Groth16 and KZG10 verifiers to verify the zkSNARK proofs coming from
|
|
Nova+CycleFold folding.
|
|
*/
|
|
|
|
|
|
/* =============================== */
|
|
/* KZG10 verifier methods */
|
|
{{ kzg10_verifier }}
|
|
|
|
/* =============================== */
|
|
/* Groth16 verifier methods */
|
|
{{ groth16_verifier }}
|
|
|
|
|
|
/* =============================== */
|
|
/* Nova+CycleFold Decider verifier */
|
|
/**
|
|
* @notice Computes the decomposition of a `uint256` into num_limbs limbs of bits_per_limb bits each.
|
|
* @dev Compatible with sonobe::folding-schemes::folding::circuits::nonnative::nonnative_field_to_field_elements.
|
|
*/
|
|
library LimbsDecomposition {
|
|
function decompose(uint256 x) internal pure returns (uint256[{{num_limbs}}] memory) {
|
|
uint256[{{num_limbs}}] memory limbs;
|
|
for (uint8 i = 0; i < {{num_limbs}}; i++) {
|
|
limbs[i] = (x >> ({{bits_per_limb}} * i)) & ((1 << {{bits_per_limb}}) - 1);
|
|
}
|
|
return limbs;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* @author PSE & 0xPARC
|
|
* @title NovaDecider contract, for verifying Nova IVC SNARK proofs.
|
|
* @dev This is an askama template which, when templated, features a Groth16 and KZG10 verifiers from which this contract inherits.
|
|
*/
|
|
contract NovaDecider is Groth16Verifier, KZG10Verifier {
|
|
/**
|
|
* @notice Computes the linear combination of a and b with r as the coefficient.
|
|
* @dev All ops are done mod the BN254 scalar field prime
|
|
*/
|
|
function rlc(uint256 a, uint256 r, uint256 b) internal pure returns (uint256 result) {
|
|
assembly {
|
|
result := addmod(a, mulmod(r, b, BN254_SCALAR_FIELD), BN254_SCALAR_FIELD)
|
|
}
|
|
}
|
|
|
|
/**
|
|
* @notice Verifies a nova cyclefold proof consisting of two KZG proofs and of a groth16 proof.
|
|
* @dev The selector of this function is "dynamic", since it depends on `z_len`.
|
|
*/
|
|
function verifyNovaProof(
|
|
// inputs are grouped to prevent errors due stack too deep
|
|
uint256[{{ 1 + z_len * 2 }}] calldata i_z0_zi, // [i, z0, zi] where |z0| == |zi|
|
|
uint256[4] calldata U_i_cmW_U_i_cmE, // [U_i_cmW[2], U_i_cmE[2]]
|
|
uint256[3] calldata U_i_u_u_i_u_r, // [U_i_u, u_i_u, r]
|
|
uint256[4] calldata U_i_x_u_i_cmW, // [U_i_x[2], u_i_cmW[2]]
|
|
uint256[4] calldata u_i_x_cmT, // [u_i_x[2], cmT[2]]
|
|
uint256[2] calldata pA, // groth16
|
|
uint256[2][2] calldata pB, // groth16
|
|
uint256[2] calldata pC, // groth16
|
|
uint256[4] calldata challenge_W_challenge_E_kzg_evals, // [challenge_W, challenge_E, eval_W, eval_E]
|
|
uint256[2][2] calldata kzg_proof // [proof_W, proof_E]
|
|
) public view returns (bool) {
|
|
|
|
require(i_z0_zi[0] >= 2, "Folding: the number of folded steps should be at least 2");
|
|
|
|
// from gamma_abc_len, we subtract 1.
|
|
uint256[{{ public_inputs_len - 1 }}] memory public_inputs;
|
|
|
|
public_inputs[0] = i_z0_zi[0];
|
|
|
|
for (uint i = 0; i < {{ z_len * 2 }}; i++) {
|
|
public_inputs[1 + i] = i_z0_zi[1 + i];
|
|
}
|
|
|
|
{
|
|
// U_i.u + r * u_i.u
|
|
uint256 u = rlc(U_i_u_u_i_u_r[0], U_i_u_u_i_u_r[2], U_i_u_u_i_u_r[1]);
|
|
// U_i.x + r * u_i.x
|
|
uint256 x0 = rlc(U_i_x_u_i_cmW[0], U_i_u_u_i_u_r[2], u_i_x_cmT[0]);
|
|
uint256 x1 = rlc(U_i_x_u_i_cmW[1], U_i_u_u_i_u_r[2], u_i_x_cmT[1]);
|
|
|
|
public_inputs[{{ z_len * 2 + 1 }}] = u;
|
|
public_inputs[{{ z_len * 2 + 2 }}] = x0;
|
|
public_inputs[{{ z_len * 2 + 3 }}] = x1;
|
|
}
|
|
|
|
{
|
|
// U_i.cmE + r * u_i.cmT
|
|
uint256[2] memory mulScalarPoint = super.mulScalar([u_i_x_cmT[2], u_i_x_cmT[3]], U_i_u_u_i_u_r[2]);
|
|
uint256[2] memory cmE = super.add([U_i_cmW_U_i_cmE[2], U_i_cmW_U_i_cmE[3]], mulScalarPoint);
|
|
|
|
{
|
|
uint256[{{num_limbs}}] memory cmE_x_limbs = LimbsDecomposition.decompose(cmE[0]);
|
|
uint256[{{num_limbs}}] memory cmE_y_limbs = LimbsDecomposition.decompose(cmE[1]);
|
|
|
|
for (uint8 k = 0; k < {{num_limbs}}; k++) {
|
|
public_inputs[{{ z_len * 2 + 4 }} + k] = cmE_x_limbs[k];
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs }} + k] = cmE_y_limbs[k];
|
|
}
|
|
}
|
|
|
|
require(this.check(cmE, kzg_proof[1], challenge_W_challenge_E_kzg_evals[1], challenge_W_challenge_E_kzg_evals[3]), "KZG: verifying proof for challenge E failed");
|
|
}
|
|
|
|
{
|
|
// U_i.cmW + r * u_i.cmW
|
|
uint256[2] memory mulScalarPoint = super.mulScalar([U_i_x_u_i_cmW[2], U_i_x_u_i_cmW[3]], U_i_u_u_i_u_r[2]);
|
|
uint256[2] memory cmW = super.add([U_i_cmW_U_i_cmE[0], U_i_cmW_U_i_cmE[1]], mulScalarPoint);
|
|
|
|
{
|
|
uint256[{{num_limbs}}] memory cmW_x_limbs = LimbsDecomposition.decompose(cmW[0]);
|
|
uint256[{{num_limbs}}] memory cmW_y_limbs = LimbsDecomposition.decompose(cmW[1]);
|
|
|
|
for (uint8 k = 0; k < {{num_limbs}}; k++) {
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 2 }} + k] = cmW_x_limbs[k];
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 3 }} + k] = cmW_y_limbs[k];
|
|
}
|
|
}
|
|
|
|
require(this.check(cmW, kzg_proof[0], challenge_W_challenge_E_kzg_evals[0], challenge_W_challenge_E_kzg_evals[2]), "KZG: verifying proof for challenge W failed");
|
|
}
|
|
|
|
{
|
|
// add challenges
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 4 }}] = challenge_W_challenge_E_kzg_evals[0];
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 4 + 1 }}] = challenge_W_challenge_E_kzg_evals[1];
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 4 + 2 }}] = challenge_W_challenge_E_kzg_evals[2];
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 4 + 3 }}] = challenge_W_challenge_E_kzg_evals[3];
|
|
|
|
uint256[{{num_limbs}}] memory cmT_x_limbs;
|
|
uint256[{{num_limbs}}] memory cmT_y_limbs;
|
|
|
|
cmT_x_limbs = LimbsDecomposition.decompose(u_i_x_cmT[2]);
|
|
cmT_y_limbs = LimbsDecomposition.decompose(u_i_x_cmT[3]);
|
|
|
|
for (uint8 k = 0; k < {{num_limbs}}; k++) {
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 4 }} + 4 + k] = cmT_x_limbs[k];
|
|
public_inputs[{{ z_len * 2 + 4 + num_limbs * 5}} + 4 + k] = cmT_y_limbs[k];
|
|
}
|
|
|
|
// last element of the groth16 proof's public inputs is `r`
|
|
public_inputs[{{ public_inputs_len - 2 }}] = U_i_u_u_i_u_r[2];
|
|
|
|
bool success_g16 = this.verifyProof(pA, pB, pC, public_inputs);
|
|
require(success_g16 == true, "Groth16: verifying proof failed");
|
|
}
|
|
|
|
return(true);
|
|
}
|
|
}
|