zkSNARK library implementation in Go
Succinct Non-Interactive Zero Knowledge for a von Neumann Architecture
, Eli Ben-Sasson, Alessandro Chiesa, Eran Tromer, Madars Virza https://eprint.iacr.org/2013/879.pdfPinocchio: Nearly practical verifiable computation
, Bryan Parno, Craig Gentry, Jon Howell, Mariana Raykova https://eprint.iacr.org/2013/279.pdfImplementation of the zkSNARK Pinocchio protocol from scratch in Go to understand the concepts. Do not use in production.
Not finished, implementing this in my free time to understand it better, so I don't have much time.
Current implementation status:
Having a circuit file test.circuit
:
func test(private s0, public s1):
s2 = s0 * s0
s3 = s2 * s0
s4 = s3 + s0
s5 = s4 + 5
equals(s1, s5)
out = 1 * 1
And a private inputs file privateInputs.json
[
3
]
And a public inputs file publicInputs.json
[
35
]
In the command line, execute:
> ./go-snark-cli compile test.circuit
This will output the compiledcircuit.json
file.
Having the compiledcircuit.json
, now we can generate the TrustedSetup
:
> ./go-snark-cli trustedsetup
This will create the file trustedsetup.json
with the TrustedSetup data, and also a toxic.json
file, with the parameters to delete from the Trusted Setup
.
Assumming that we have the compiledcircuit.json
, trustedsetup.json
, privateInputs.json
and the publicInputs.json
we can now generate the Proofs
with the following command:
> ./go-snark-cli genproofs
This will store the file proofs.json
, that contains all the SNARK proofs.
Having the proofs.json
, compiledcircuit.json
, trustedsetup.json
publicInputs.json
files, we can now verify the Pairings
of the proofs, in order to verify the proofs.
> ./go-snark-cli verify
This will return a true
if the proofs are verified, or a false
if the proofs are not verified.
Warning: not finished.
Example:
// compile circuit and get the R1CS
flatCode := `
func test(private s0, public s1):
s2 = s0 * s0
s3 = s2 * s0
s4 = s3 + s0
s5 = s4 + 5
equals(s1, s5)
out = 1 * 1
`
// parse the code
parser := circuitcompiler.NewParser(strings.NewReader(flatCode))
circuit, err := parser.Parse()
assert.Nil(t, err)
fmt.Println(circuit)
b3 := big.NewInt(int64(3))
privateInputs := []*big.Int{b3}
b35 := big.NewInt(int64(35))
publicSignals := []*big.Int{b35}
// witness
w, err := circuit.CalculateWitness(privateInputs, publicSignals)
assert.Nil(t, err)
fmt.Println("witness", w)
// now we have the witness:
// w = [1 35 3 9 27 30 35 1]
// flat code to R1CS
fmt.Println("generating R1CS from flat code")
a, b, c := circuit.GenerateR1CS()
/*
now we have the R1CS from the circuit:
a: [[0 0 1 0 0 0 0 0] [0 0 0 1 0 0 0 0] [0 0 1 0 1 0 0 0] [5 0 0 0 0 1 0 0] [0 0 0 0 0 0 1 0] [0 1 0 0 0 0 0 0] [1 0 0 0 0 0 0 0]]
b: [[0 0 1 0 0 0 0 0] [0 0 1 0 0 0 0 0] [1 0 0 0 0 0 0 0] [1 0 0 0 0 0 0 0] [1 0 0 0 0 0 0 0] [1 0 0 0 0 0 0 0] [1 0 0 0 0 0 0 0]]
c: [[0 0 0 1 0 0 0 0] [0 0 0 0 1 0 0 0] [0 0 0 0 0 1 0 0] [0 0 0 0 0 0 1 0] [0 1 0 0 0 0 0 0] [0 0 0 0 0 0 1 0] [0 0 0 0 0 0 0 1]]
*/
alphas, betas, gammas, _ := snark.Utils.PF.R1CSToQAP(a, b, c)
ax, bx, cx, px := Utils.PF.CombinePolynomials(w, alphas, betas, gammas)
// calculate trusted setup
setup, err := GenerateTrustedSetup(len(w), *circuit, alphas, betas, gammas)
hx := Utils.PF.DivisorPolynomial(px, setup.Pk.Z)
proof, err := GenerateProofs(*circuit, setup, w, px)
b35Verif := big.NewInt(int64(35))
publicSignalsVerif := []*big.Int{b35Verif}
assert.True(t, VerifyProof(*circuit, setup, proof, publicSignalsVerif, true))
go test ./... -v
Thanks to @jbaylina, @bellesmarta, @adriamb for their explanations that helped to understand this a little bit. Also thanks to @vbuterin for all the published articles explaining the zkSNARKs.