eth2.0-specs/specs/_features/eip7594/polynomial-commitments-samp...

695 lines
25 KiB
Markdown
Raw Normal View History

wip wip Add `TARGET_NUMBER_OF_PEERS` Add networking spec draft fix simplification Rename `DoYouHave` to `GetCustodyStatus` Add DataLineSidecar design Apply suggestions from code review Co-authored-by: dankrad <mail@dankradfeist.de> Co-authored-by: danny <dannyjryan@gmail.com> Revamp after reviews and discussion Remove `CustodyStatus` minor fix Change`DataColumn` to `List[DataCell, MAX_BLOBS_PER_BLOCK]` Move folder Replace `DataColumnByRootAndIndex` with `DataColumnSidecarByRoot` message. Add extended data description Remove `DataRow` Apply suggestions from @jacobkaufmann code review Co-authored-by: Jacob Kaufmann <jacobkaufmann18@gmail.com> Represent matrix in `BLSFieldElement` form Add `assert time >= store.time` to `on_tick` Revert the spec. Only handle it in tests Remove extra tick cleanup leftover Add randomized block cases Specify RPC byRoot blocks-sidecars elegibility fix typo Update specs/phase0/p2p-interface.md Co-authored-by: Mikhail Kalinin <noblesse.knight@gmail.com> Update specs/deneb/p2p-interface.md Co-authored-by: Mikhail Kalinin <noblesse.knight@gmail.com> add failed on_block condition rephrase Update specs/phase0/p2p-interface.md Co-authored-by: Mikhail Kalinin <noblesse.knight@gmail.com> apply suggestion Update specs/deneb/p2p-interface.md Co-authored-by: danny <dannyjryan@gmail.com> Update specs/deneb/p2p-interface.md Co-authored-by: danny <dannyjryan@gmail.com> remove the last consider from on_block to state_transition simplify and add a new rule Update specs/phase0/p2p-interface.md Co-authored-by: Mikhail Kalinin <noblesse.knight@gmail.com> Update specs/deneb/p2p-interface.md Co-authored-by: Mikhail Kalinin <noblesse.knight@gmail.com> Update specs/deneb/p2p-interface.md Co-authored-by: danny <dannyjryan@gmail.com> remove gossip failure rules Apply suggestions from code review bump version to v1.4.0-beta.5 Move `blob_sidecar_{subnet_id}` to `Blob subnets` section Misc minor fix Add linter support Add column subnet validation. Split `verify_column_sidecar` into two functions Fix `get_data_column_sidecars` by using `compute_samples_and_proofs` Apply suggestions from code review Co-authored-by: danny <dannyjryan@gmail.com> Do not assign row custody Apply suggestions from code review Co-authored-by: danny <dannyjryan@gmail.com> Revamp reconstruction section Use depth as the primary preset for inclusion proof. Fix `get_data_column_sidecars` and add tests for merkle proof Change `SAMPLES_PER_SLOT` to 8 and add tests (requirement TBD) Apply PR feedback from @ppopth and @jtraglia Fix `get_data_column_sidecars` Co-authored-by: Pop Chunhapanya <haxx.pop@gmail.com> Apply suggestions from code review Co-authored-by: Pop Chunhapanya <haxx.pop@gmail.com> Apply suggestions from code review Co-authored-by: fradamt <104826920+fradamt@users.noreply.github.com> Co-authored-by: Jacob Kaufmann <jacobkaufmann18@gmail.com> Fix `get_data_column_sidecars` and `get_custody_lines` Apply suggestions from code review Co-authored-by: Jacob Kaufmann <jacobkaufmann18@gmail.com> Enhance tests fix typo Co-authored-by: fradamt <104826920+fradamt@users.noreply.github.com> Remove `epoch` from `get_custody_lines` fix fix
2023-11-14 09:04:42 +00:00
# EIP-7594 -- Polynomial Commitments
## Table of contents
<!-- TOC -->
<!-- START doctoc generated TOC please keep comment here to allow auto update -->
<!-- DON'T EDIT THIS SECTION, INSTEAD RE-RUN doctoc TO UPDATE -->
- [Introduction](#introduction)
- [Custom types](#custom-types)
- [Constants](#constants)
- [Preset](#preset)
2023-12-23 14:52:39 +00:00
- [Cells](#cells)
- [Helper functions](#helper-functions)
2024-01-15 16:27:55 +00:00
- [BLS12-381 helpers](#bls12-381-helpers)
- [`cell_to_coset_evals`](#cell_to_coset_evals)
- [`coset_evals_to_cell`](#coset_evals_to_cell)
- [Linear combinations](#linear-combinations)
- [`g2_lincomb`](#g2_lincomb)
- [FFTs](#ffts)
- [`_fft_field`](#_fft_field)
- [`fft_field`](#fft_field)
- [Polynomials in coefficient form](#polynomials-in-coefficient-form)
- [`polynomial_eval_to_coeff`](#polynomial_eval_to_coeff)
- [`add_polynomialcoeff`](#add_polynomialcoeff)
- [`neg_polynomialcoeff`](#neg_polynomialcoeff)
- [`multiply_polynomialcoeff`](#multiply_polynomialcoeff)
- [`divide_polynomialcoeff`](#divide_polynomialcoeff)
- [`shift_polynomialcoeff`](#shift_polynomialcoeff)
- [`interpolate_polynomialcoeff`](#interpolate_polynomialcoeff)
- [`vanishing_polynomialcoeff`](#vanishing_polynomialcoeff)
- [`evaluate_polynomialcoeff`](#evaluate_polynomialcoeff)
- [KZG multiproofs](#kzg-multiproofs)
- [`compute_kzg_proof_multi_impl`](#compute_kzg_proof_multi_impl)
- [`verify_kzg_proof_multi_impl`](#verify_kzg_proof_multi_impl)
2023-12-23 14:52:39 +00:00
- [Cell cosets](#cell-cosets)
- [`coset_for_cell`](#coset_for_cell)
- [Cells](#cells-1)
- [Cell computation](#cell-computation)
- [`compute_cells_and_proofs`](#compute_cells_and_proofs)
- [`compute_cells`](#compute_cells)
- [Cell verification](#cell-verification)
- [`verify_cell_proof`](#verify_cell_proof)
- [`verify_cell_proof_batch`](#verify_cell_proof_batch)
- [Reconstruction](#reconstruction)
- [`construct_vanishing_polynomial`](#construct_vanishing_polynomial)
- [`recover_shifted_data`](#recover_shifted_data)
- [`recover_original_data`](#recover_original_data)
2024-04-19 11:50:22 +00:00
- [`recover_all_cells`](#recover_all_cells)
<!-- END doctoc generated TOC please keep comment here to allow auto update -->
<!-- /TOC -->
## Introduction
This document extends [polynomial-commitments.md](polynomial-commitments.md) with the functions required for data availability sampling (DAS). It is not part of the core Deneb spec but an extension that can be optionally implemented to allow nodes to reduce their load using DAS.
For any KZG library extended to support DAS, functions flagged as "Public method" MUST be provided by the underlying KZG library as public functions. All other functions are private functions used internally by the KZG library.
Public functions MUST accept raw bytes as input and perform the required cryptographic normalization before invoking any internal functions.
## Custom types
| Name | SSZ equivalent | Description |
| - | - | - |
| `PolynomialCoeff` | `List[BLSFieldElement, FIELD_ELEMENTS_PER_EXT_BLOB]` | A polynomial in coefficient form |
| `Coset` | `Vector[BLSFieldElement, FIELD_ELEMENTS_PER_CELL]` | The evaluation domain of a cell |
| `CosetEvals` | `Vector[BLSFieldElement, FIELD_ELEMENTS_PER_CELL]` | The internal representation of a cell (the evaluations over its Coset) |
| `Cell` | `ByteVector[BYTES_PER_FIELD_ELEMENT * FIELD_ELEMENTS_PER_CELL]` | The unit of blob data that can come with its own KZG proof |
| `CellID` | `uint64` | Cell identifier |
| `RowIndex` | `uint64` | Row identifier |
| `ColumnIndex` | `uint64` | Column identifier |
## Constants
| Name | Value | Notes |
| - | - | - |
## Preset
2023-12-23 14:52:39 +00:00
### Cells
Cells are the smallest unit of blob data that can come with their own KZG proofs. Samples can be constructed from one or several cells (e.g. an individual cell or line).
| Name | Value | Description |
| - | - | - |
| `FIELD_ELEMENTS_PER_EXT_BLOB` | `2 * FIELD_ELEMENTS_PER_BLOB` | Number of field elements in a Reed-Solomon extended blob |
2023-12-23 14:44:58 +00:00
| `FIELD_ELEMENTS_PER_CELL` | `uint64(64)` | Number of field elements in a cell |
| `BYTES_PER_CELL` | `FIELD_ELEMENTS_PER_CELL * BYTES_PER_FIELD_ELEMENT` | The number of bytes in a cell |
2024-04-18 16:38:04 +00:00
| `CELLS_PER_EXT_BLOB` | `FIELD_ELEMENTS_PER_EXT_BLOB // FIELD_ELEMENTS_PER_CELL` | The number of cells in an extended blob |
| `RANDOM_CHALLENGE_KZG_CELL_BATCH_DOMAIN` | `b'RCKZGCBATCH__V1_'` |
## Helper functions
2024-01-15 16:27:55 +00:00
### BLS12-381 helpers
#### `cell_to_coset_evals`
2024-01-15 16:27:55 +00:00
```python
def cell_to_coset_evals(cell: Cell) -> CosetEvals:
2024-01-15 16:27:55 +00:00
"""
Convert an untrusted ``Cell`` into a trusted ``CosetEvals``.
2024-01-15 16:27:55 +00:00
"""
evals = []
for i in range(FIELD_ELEMENTS_PER_CELL):
start = i * BYTES_PER_FIELD_ELEMENT
end = (i + 1) * BYTES_PER_FIELD_ELEMENT
value = bytes_to_bls_field(cell[start:end])
evals.append(value)
return CosetEvals(evals)
```
#### `coset_evals_to_cell`
```python
def coset_evals_to_cell(coset_evals: CosetEvals) -> Cell:
"""
Convert a trusted ``CosetEval`` into an untrusted ``Cell``.
"""
cell = []
for i in range(FIELD_ELEMENTS_PER_CELL):
cell += bls_field_to_bytes(coset_evals[i])
return Cell(cell)
2024-01-15 16:27:55 +00:00
```
### Linear combinations
#### `g2_lincomb`
```python
def g2_lincomb(points: Sequence[G2Point], scalars: Sequence[BLSFieldElement]) -> Bytes96:
"""
BLS multiscalar multiplication in G2. This function can be optimized using Pippenger's algorithm and variants.
"""
assert len(points) == len(scalars)
2024-04-23 12:38:27 +00:00
points_g2 = []
for point in points:
points_g2.append(bls.bytes96_to_G2(point))
result = bls.g2_multi_exp(points_g2, scalars)
return Bytes96(bls.G2_to_bytes96(result))
```
### FFTs
#### `_fft_field`
```python
2024-01-05 12:31:09 +00:00
def _fft_field(vals: Sequence[BLSFieldElement],
roots_of_unity: Sequence[BLSFieldElement]) -> Sequence[BLSFieldElement]:
if len(vals) == 1:
2023-12-23 14:44:58 +00:00
return vals
L = _fft_field(vals[::2], roots_of_unity[::2])
R = _fft_field(vals[1::2], roots_of_unity[::2])
2024-01-05 12:31:09 +00:00
o = [BLSFieldElement(0) for _ in vals]
for i, (x, y) in enumerate(zip(L, R)):
2024-01-05 12:31:09 +00:00
y_times_root = (int(y) * int(roots_of_unity[i])) % BLS_MODULUS
o[i] = BLSFieldElement((int(x) + y_times_root) % BLS_MODULUS)
o[i + len(L)] = BLSFieldElement((int(x) - y_times_root + BLS_MODULUS) % BLS_MODULUS)
return o
```
#### `fft_field`
```python
2024-01-05 12:31:09 +00:00
def fft_field(vals: Sequence[BLSFieldElement],
roots_of_unity: Sequence[BLSFieldElement],
inv: bool=False) -> Sequence[BLSFieldElement]:
if inv:
# Inverse FFT
invlen = pow(len(vals), BLS_MODULUS - 2, BLS_MODULUS)
2024-01-05 12:31:09 +00:00
return [BLSFieldElement((int(x) * invlen) % BLS_MODULUS)
for x in _fft_field(vals, list(roots_of_unity[0:1]) + list(roots_of_unity[:0:-1]))]
else:
# Regular FFT
return _fft_field(vals, roots_of_unity)
```
### Polynomials in coefficient form
#### `polynomial_eval_to_coeff`
```python
def polynomial_eval_to_coeff(polynomial: Polynomial) -> PolynomialCoeff:
"""
Interpolates a polynomial (given in evaluation form) to a polynomial in coefficient form.
"""
roots_of_unity = compute_roots_of_unity(FIELD_ELEMENTS_PER_BLOB)
polynomial_coeff = fft_field(bit_reversal_permutation(list(polynomial)), roots_of_unity, inv=True)
return polynomial_coeff
```
#### `add_polynomialcoeff`
```python
def add_polynomialcoeff(a: PolynomialCoeff, b: PolynomialCoeff) -> PolynomialCoeff:
"""
Sum the coefficient form polynomials ``a`` and ``b``.
"""
a, b = (a, b) if len(a) >= len(b) else (b, a)
2024-01-16 08:51:05 +00:00
length_a = len(a)
length_b = len(b)
return [(a[i] + (b[i] if i < length_b else 0)) % BLS_MODULUS for i in range(length_a)]
```
#### `neg_polynomialcoeff`
```python
def neg_polynomialcoeff(a: PolynomialCoeff) -> PolynomialCoeff:
"""
Negative of coefficient form polynomial ``a``
"""
return [(BLS_MODULUS - x) % BLS_MODULUS for x in a]
```
#### `multiply_polynomialcoeff`
```python
def multiply_polynomialcoeff(a: PolynomialCoeff, b: PolynomialCoeff) -> PolynomialCoeff:
"""
Multiplies the coefficient form polynomials ``a`` and ``b``
"""
assert len(a) + len(b) <= FIELD_ELEMENTS_PER_EXT_BLOB
r = [0]
for power, coef in enumerate(a):
summand = [0] * power + [int(coef) * int(x) % BLS_MODULUS for x in b]
r = add_polynomialcoeff(r, summand)
return r
```
#### `divide_polynomialcoeff`
```python
def divide_polynomialcoeff(a: PolynomialCoeff, b: PolynomialCoeff) -> PolynomialCoeff:
"""
Long polynomial division for two coefficient form polynomials ``a`` and ``b``
"""
a = a.copy() # Make a copy since `a` is passed by reference
o = []
apos = len(a) - 1
bpos = len(b) - 1
diff = apos - bpos
while diff >= 0:
quot = div(a[apos], b[bpos])
o.insert(0, quot)
for i in range(bpos, -1, -1):
a[diff + i] = (int(a[diff + i]) - int(b[i] + BLS_MODULUS) * int(quot)) % BLS_MODULUS
apos -= 1
diff -= 1
return [x % BLS_MODULUS for x in o]
```
#### `shift_polynomialcoeff`
```python
def shift_polynomialcoeff(polynomial_coeff: PolynomialCoeff, factor: BLSFieldElement) -> PolynomialCoeff:
"""
Shift the evaluation of a polynomial in coefficient form by factor.
This results in a new polynomial g(x) = f(factor * x)
"""
factor_power = 1
inv_factor = pow(int(factor), BLS_MODULUS - 2, BLS_MODULUS)
o = []
for p in polynomial_coeff:
o.append(int(p) * factor_power % BLS_MODULUS)
factor_power = factor_power * inv_factor % BLS_MODULUS
return o
```
#### `interpolate_polynomialcoeff`
```python
def interpolate_polynomialcoeff(xs: Sequence[BLSFieldElement], ys: Sequence[BLSFieldElement]) -> PolynomialCoeff:
"""
Lagrange interpolation: Finds the lowest degree polynomial that takes the value ``ys[i]`` at ``x[i]``
for all i.
Outputs a coefficient form polynomial. Leading coefficients may be zero.
"""
assert len(xs) == len(ys)
r = [0]
for i in range(len(xs)):
summand = [ys[i]]
for j in range(len(ys)):
if j != i:
weight_adjustment = bls_modular_inverse(int(xs[i]) - int(xs[j]))
summand = multiply_polynomialcoeff(
summand, [((BLS_MODULUS - int(weight_adjustment)) * int(xs[j])) % BLS_MODULUS, weight_adjustment]
)
r = add_polynomialcoeff(r, summand)
2024-01-15 15:34:28 +00:00
return r
```
#### `vanishing_polynomialcoeff`
```python
def vanishing_polynomialcoeff(xs: Sequence[BLSFieldElement]) -> PolynomialCoeff:
"""
Compute the vanishing polynomial on ``xs`` (in coefficient form)
"""
p = [1]
for x in xs:
p = multiply_polynomialcoeff(p, [-int(x) + BLS_MODULUS, 1])
return p
```
#### `evaluate_polynomialcoeff`
```python
def evaluate_polynomialcoeff(polynomial_coeff: PolynomialCoeff, z: BLSFieldElement) -> BLSFieldElement:
"""
Evaluate a coefficient form polynomial at ``z`` using Horner's schema
"""
y = 0
for coef in polynomial_coeff[::-1]:
2024-01-05 12:31:09 +00:00
y = (int(y) * int(z) + int(coef)) % BLS_MODULUS
return BLSFieldElement(y % BLS_MODULUS)
```
### KZG multiproofs
Extended KZG functions for multiproofs
#### `compute_kzg_proof_multi_impl`
```python
2023-12-12 16:25:30 +00:00
def compute_kzg_proof_multi_impl(
polynomial_coeff: PolynomialCoeff,
zs: Coset) -> Tuple[KZGProof, CosetEvals]:
"""
2024-04-18 16:52:33 +00:00
Compute a KZG multi-evaluation proof for a set of `k` points.
This is done by committing to the following quotient polynomial:
2024-04-22 08:57:58 +00:00
Q(X) = f(X) - I(X) / Z(X)
2024-04-18 16:52:33 +00:00
Where:
2024-04-22 08:57:58 +00:00
- I(X) is the degree `k-1` polynomial that agrees with f(x) at all `k` points
2024-04-18 16:52:33 +00:00
- Z(X) is the degree `k` polynomial that evaluates to zero on all `k` points
2024-04-22 08:57:58 +00:00
We further note that since the degree of I(X) is less than the degree of Z(X),
the computation can be simplified in monomial form to Q(X) = f(X) / Z(X)
"""
# For all points, compute the evaluation of those points
ys = [evaluate_polynomialcoeff(polynomial_coeff, z) for z in zs]
2024-04-18 16:52:33 +00:00
# Compute Z(X)
denominator_poly = vanishing_polynomialcoeff(zs)
2024-04-18 16:52:33 +00:00
# Compute the quotient polynomial directly in monomial form
quotient_polynomial = divide_polynomialcoeff(polynomial_coeff, denominator_poly)
return KZGProof(g1_lincomb(KZG_SETUP_G1_MONOMIAL[:len(quotient_polynomial)], quotient_polynomial)), ys
```
#### `verify_kzg_proof_multi_impl`
```python
def verify_kzg_proof_multi_impl(commitment: KZGCommitment,
zs: Coset,
ys: CosetEvals,
proof: KZGProof) -> bool:
"""
2024-04-19 18:19:17 +00:00
Verify a KZG multi-evaluation proof for a set of `k` points.
This is done by checking if the following equation holds:
2024-04-22 08:57:58 +00:00
Q(x) Z(x) = f(X) - I(X)
2024-04-19 18:19:17 +00:00
Where:
2024-04-22 08:57:58 +00:00
f(X) is the polynomial that we want to verify opens at `k` points to `k` values
2024-04-19 18:19:17 +00:00
Q(X) is the quotient polynomial computed by the prover
2024-04-22 08:57:58 +00:00
I(X) is the degree k-1 polynomial that evaluates to `ys` at all `zs`` points
2024-04-19 18:19:17 +00:00
Z(X) is the polynomial that evaluates to zero on all `k` points
The verifier receives the commitments to Q(X) and f(X), so they check the equation
holds by using the following pairing equation:
2024-04-22 08:57:58 +00:00
e([Q(X)]_1, [Z(X)]_2) == e([f(X)]_1 - [I(X)]_1, [1]_2)
"""
assert len(zs) == len(ys)
2024-04-19 18:19:17 +00:00
# Compute [Z(X)]_2
zero_poly = g2_lincomb(KZG_SETUP_G2_MONOMIAL[:len(zs) + 1], vanishing_polynomialcoeff(zs))
2024-04-22 08:57:58 +00:00
# Compute [I(X)]_1
interpolated_poly = g1_lincomb(KZG_SETUP_G1_MONOMIAL[:len(zs)], interpolate_polynomialcoeff(zs, ys))
2023-12-12 16:25:30 +00:00
return (bls.pairing_check([
[bls.bytes48_to_G1(proof), bls.bytes96_to_G2(zero_poly)],
[
bls.add(bls.bytes48_to_G1(commitment), bls.neg(bls.bytes48_to_G1(interpolated_poly))),
bls.neg(bls.bytes96_to_G2(KZG_SETUP_G2_MONOMIAL[0])),
2023-12-12 16:25:30 +00:00
],
]))
```
2023-12-23 14:52:39 +00:00
### Cell cosets
2023-12-23 14:52:39 +00:00
#### `coset_for_cell`
```python
def coset_for_cell(cell_id: CellID) -> Coset:
"""
2023-12-23 14:52:39 +00:00
Get the coset for a given ``cell_id``
"""
2024-04-18 16:38:04 +00:00
assert cell_id < CELLS_PER_EXT_BLOB
roots_of_unity_brp = bit_reversal_permutation(
compute_roots_of_unity(FIELD_ELEMENTS_PER_EXT_BLOB)
)
return Coset(roots_of_unity_brp[FIELD_ELEMENTS_PER_CELL * cell_id:FIELD_ELEMENTS_PER_CELL * (cell_id + 1)])
```
2023-12-23 14:52:39 +00:00
## Cells
2023-12-23 14:52:39 +00:00
### Cell computation
2023-12-23 14:52:39 +00:00
#### `compute_cells_and_proofs`
```python
2023-12-23 14:52:39 +00:00
def compute_cells_and_proofs(blob: Blob) -> Tuple[
2024-04-18 16:38:04 +00:00
Vector[Cell, CELLS_PER_EXT_BLOB],
Vector[KZGProof, CELLS_PER_EXT_BLOB]]:
"""
2024-04-18 17:11:41 +00:00
Compute all the cell proofs for an extended blob. This is an inefficient O(n^2) algorithm,
for performant implementation the FK20 algorithm that runs in O(n log n) should be
used instead.
2023-12-04 01:26:07 +00:00
Public method.
"""
polynomial = blob_to_polynomial(blob)
polynomial_coeff = polynomial_eval_to_coeff(polynomial)
2023-12-23 14:52:39 +00:00
cells = []
proofs = []
2024-04-18 16:38:04 +00:00
for i in range(CELLS_PER_EXT_BLOB):
2023-12-23 14:52:39 +00:00
coset = coset_for_cell(i)
proof, ys = compute_kzg_proof_multi_impl(polynomial_coeff, coset)
cells.append(coset_evals_to_cell(ys))
proofs.append(proof)
2023-12-23 14:52:39 +00:00
return cells, proofs
```
2023-12-23 14:52:39 +00:00
#### `compute_cells`
```python
2024-04-18 16:38:04 +00:00
def compute_cells(blob: Blob) -> Vector[Cell, CELLS_PER_EXT_BLOB]:
2023-12-04 01:26:07 +00:00
"""
2024-04-18 17:12:15 +00:00
Compute the cell data for an extended blob (without computing the proofs).
2023-12-04 01:26:07 +00:00
Public method.
"""
polynomial = blob_to_polynomial(blob)
polynomial_coeff = polynomial_eval_to_coeff(polynomial)
2024-01-15 15:34:28 +00:00
extended_data = fft_field(polynomial_coeff + [0] * FIELD_ELEMENTS_PER_BLOB,
compute_roots_of_unity(FIELD_ELEMENTS_PER_EXT_BLOB))
extended_data_rbo = bit_reversal_permutation(extended_data)
cells = []
for cell_id in range(CELLS_PER_EXT_BLOB):
start = cell_id * FIELD_ELEMENTS_PER_CELL
end = (cell_id + 1) * FIELD_ELEMENTS_PER_CELL
cells.append(coset_evals_to_cell(extended_data_rbo[start:end]))
return cells
```
2023-12-23 14:52:39 +00:00
### Cell verification
2023-12-23 14:52:39 +00:00
#### `verify_cell_proof`
```python
2024-01-15 16:27:55 +00:00
def verify_cell_proof(commitment_bytes: Bytes48,
2024-01-17 12:18:03 +00:00
cell_id: CellID,
cell: Cell,
2024-01-17 12:18:03 +00:00
proof_bytes: Bytes48) -> bool:
"""
2023-12-23 14:52:39 +00:00
Check a cell proof
2023-12-04 01:26:07 +00:00
2023-12-23 14:44:58 +00:00
Public method.
"""
2023-12-23 14:52:39 +00:00
coset = coset_for_cell(cell_id)
2024-01-15 16:27:55 +00:00
return verify_kzg_proof_multi_impl(
bytes_to_kzg_commitment(commitment_bytes),
coset,
cell_to_coset_evals(cell),
2024-01-17 12:18:03 +00:00
bytes_to_kzg_proof(proof_bytes))
```
2023-12-23 14:52:39 +00:00
#### `verify_cell_proof_batch`
```python
2024-01-15 16:27:55 +00:00
def verify_cell_proof_batch(row_commitments_bytes: Sequence[Bytes48],
row_indices: Sequence[RowIndex],
column_indices: Sequence[ColumnIndex],
cells: Sequence[Cell],
2024-01-15 16:27:55 +00:00
proofs_bytes: Sequence[Bytes48]) -> bool:
"""
2024-01-17 15:20:36 +00:00
Verify a set of cells, given their corresponding proofs and their coordinates (row_id, column_id) in the blob
matrix. The list of all commitments is also provided in row_commitments_bytes.
This function implements the naive algorithm of checking every cell
individually; an efficient algorithm can be found here:
https://ethresear.ch/t/a-universal-verification-equation-for-data-availability-sampling/13240
2023-12-04 01:26:07 +00:00
This implementation does not require randomness, but for the algorithm that
requires it, `RANDOM_CHALLENGE_KZG_CELL_BATCH_DOMAIN` should be used to compute
the challenge value.
2023-12-04 01:26:07 +00:00
Public method.
"""
assert len(cells) == len(proofs_bytes) == len(row_indices) == len(column_indices)
# Get commitments via row IDs
2024-01-29 09:29:20 +00:00
commitments_bytes = [row_commitments_bytes[row_index] for row_index in row_indices]
2024-01-15 16:27:55 +00:00
# Get objects from bytes
commitments = [bytes_to_kzg_commitment(commitment_bytes) for commitment_bytes in commitments_bytes]
cosets_evals = [cell_to_coset_evals(cell) for cell in cells]
2024-01-15 16:27:55 +00:00
proofs = [bytes_to_kzg_proof(proof_bytes) for proof_bytes in proofs_bytes]
return all(
verify_kzg_proof_multi_impl(commitment, coset_for_cell(column_index), coset_evals, proof)
for commitment, column_index, coset_evals, proof in zip(commitments, column_indices, cosets_evals, proofs)
)
```
## Reconstruction
### `construct_vanishing_polynomial`
```python
def construct_vanishing_polynomial(missing_cell_ids: Sequence[CellID]) -> Tuple[
Sequence[BLSFieldElement],
Sequence[BLSFieldElement]]:
"""
Given the cells that are missing from the data, compute the polynomial that vanishes at every point that
corresponds to a missing field element.
"""
# Get the small domain
2024-04-18 16:38:04 +00:00
roots_of_unity_reduced = compute_roots_of_unity(CELLS_PER_EXT_BLOB)
# Compute polynomial that vanishes at all the missing cells (over the small domain)
2024-01-05 10:33:01 +00:00
short_zero_poly = vanishing_polynomialcoeff([
2024-04-18 16:38:04 +00:00
roots_of_unity_reduced[reverse_bits(missing_cell_id, CELLS_PER_EXT_BLOB)]
for missing_cell_id in missing_cell_ids
2024-01-05 10:33:01 +00:00
])
# Extend vanishing polynomial to full domain using the closed form of the vanishing polynomial over a coset
zero_poly_coeff = [0] * FIELD_ELEMENTS_PER_EXT_BLOB
for i, coeff in enumerate(short_zero_poly):
zero_poly_coeff[i * FIELD_ELEMENTS_PER_CELL] = coeff
# Compute evaluations of the extended vanishing polynomial
zero_poly_eval = fft_field(zero_poly_coeff,
compute_roots_of_unity(FIELD_ELEMENTS_PER_EXT_BLOB))
zero_poly_eval_brp = bit_reversal_permutation(zero_poly_eval)
# Sanity check
2024-04-18 16:38:04 +00:00
for cell_id in range(CELLS_PER_EXT_BLOB):
2024-01-05 10:33:01 +00:00
start = cell_id * FIELD_ELEMENTS_PER_CELL
end = (cell_id + 1) * FIELD_ELEMENTS_PER_CELL
if cell_id in missing_cell_ids:
assert all(a == 0 for a in zero_poly_eval_brp[start:end])
else: # cell_id in cell_ids
assert all(a != 0 for a in zero_poly_eval_brp[start:end])
2024-04-20 01:16:19 +00:00
return zero_poly_coeff, zero_poly_eval
```
### `recover_shifted_data`
```python
def recover_shifted_data(cell_ids: Sequence[CellID],
cells: Sequence[Cell],
zero_poly_eval: Sequence[BLSFieldElement],
zero_poly_coeff: Sequence[BLSFieldElement],
roots_of_unity_extended: Sequence[BLSFieldElement]) -> Tuple[
Sequence[BLSFieldElement],
Sequence[BLSFieldElement],
BLSFieldElement]:
"""
Given Z(x), return polynomial Q_1(x)=(E*Z)(k*x) and Q_2(x)=Z(k*x) and k^{-1}.
"""
shift_factor = BLSFieldElement(PRIMITIVE_ROOT_OF_UNITY)
shift_inv = div(BLSFieldElement(1), shift_factor)
extended_evaluation_rbo = [0] * FIELD_ELEMENTS_PER_EXT_BLOB
for cell_id, cell in zip(cell_ids, cells):
start = cell_id * FIELD_ELEMENTS_PER_CELL
end = (cell_id + 1) * FIELD_ELEMENTS_PER_CELL
extended_evaluation_rbo[start:end] = cell
extended_evaluation = bit_reversal_permutation(extended_evaluation_rbo)
# Compute (E*Z)(x)
extended_evaluation_times_zero = [BLSFieldElement(int(a) * int(b) % BLS_MODULUS)
for a, b in zip(zero_poly_eval, extended_evaluation)]
extended_evaluations_fft = fft_field(extended_evaluation_times_zero, roots_of_unity_extended, inv=True)
# Compute (E*Z)(k*x)
shifted_extended_evaluation = shift_polynomialcoeff(extended_evaluations_fft, shift_factor)
# Compute Z(k*x)
shifted_zero_poly = shift_polynomialcoeff(zero_poly_coeff, shift_factor)
2023-12-12 16:25:30 +00:00
eval_shifted_extended_evaluation = fft_field(shifted_extended_evaluation, roots_of_unity_extended)
eval_shifted_zero_poly = fft_field(shifted_zero_poly, roots_of_unity_extended)
2024-01-15 15:34:28 +00:00
return eval_shifted_extended_evaluation, eval_shifted_zero_poly, shift_inv
```
### `recover_original_data`
```python
def recover_original_data(eval_shifted_extended_evaluation: Sequence[BLSFieldElement],
eval_shifted_zero_poly: Sequence[BLSFieldElement],
shift_inv: BLSFieldElement,
roots_of_unity_extended: Sequence[BLSFieldElement]) -> Sequence[BLSFieldElement]:
"""
Given Q_1, Q_2 and k^{-1}, compute P(x).
"""
# Compute Q_3 = Q_1(x)/Q_2(x) = P(k*x)
2023-12-12 16:25:30 +00:00
eval_shifted_reconstructed_poly = [
div(a, b)
for a, b in zip(eval_shifted_extended_evaluation, eval_shifted_zero_poly)
]
shifted_reconstructed_poly = fft_field(eval_shifted_reconstructed_poly, roots_of_unity_extended, inv=True)
2023-12-12 16:25:30 +00:00
# Unshift P(k*x) by k^{-1} to get P(x)
reconstructed_poly = shift_polynomialcoeff(shifted_reconstructed_poly, shift_inv)
reconstructed_data = bit_reversal_permutation(fft_field(reconstructed_poly, roots_of_unity_extended))
return reconstructed_data
```
2024-04-19 11:50:22 +00:00
### `recover_all_cells`
```python
def recover_all_cells(cell_ids: Sequence[CellID], cells: Sequence[Cell]) -> Sequence[Cell]:
"""
2024-04-19 11:50:22 +00:00
Recover all of the cells in the extended blob from FIELD_ELEMENTS_PER_EXT_BLOB evaluations,
half of which can be missing.
This algorithm uses FFTs to recover cells faster than using Lagrange implementation, as can be seen here:
https://ethresear.ch/t/reed-solomon-erasure-code-recovery-in-n-log-2-n-time-with-ffts/3039
A faster version thanks to Qi Zhou can be found here:
https://github.com/ethereum/research/blob/51b530a53bd4147d123ab3e390a9d08605c2cdb8/polynomial_reconstruction/polynomial_reconstruction_danksharding.py
Public method.
"""
assert len(cell_ids) == len(cells)
# Check we have enough cells to be able to perform the reconstruction
2024-04-18 16:38:04 +00:00
assert CELLS_PER_EXT_BLOB / 2 <= len(cell_ids) <= CELLS_PER_EXT_BLOB
# Check for duplicates
assert len(cell_ids) == len(set(cell_ids))
# Get the extended domain
roots_of_unity_extended = compute_roots_of_unity(FIELD_ELEMENTS_PER_EXT_BLOB)
# Convert cells to coset evals
cosets_evals = [cell_to_coset_evals(cell) for cell in cells]
2024-04-18 16:38:04 +00:00
missing_cell_ids = [cell_id for cell_id in range(CELLS_PER_EXT_BLOB) if cell_id not in cell_ids]
2024-04-20 01:16:19 +00:00
zero_poly_coeff, zero_poly_eval = construct_vanishing_polynomial(missing_cell_ids)
eval_shifted_extended_evaluation, eval_shifted_zero_poly, shift_inv = recover_shifted_data(
cell_ids,
cosets_evals,
zero_poly_eval,
zero_poly_coeff,
roots_of_unity_extended,
)
reconstructed_data = recover_original_data(
eval_shifted_extended_evaluation,
eval_shifted_zero_poly,
shift_inv,
roots_of_unity_extended,
)
for cell_id, coset_evals in zip(cell_ids, cosets_evals):
start = cell_id * FIELD_ELEMENTS_PER_CELL
end = (cell_id + 1) * FIELD_ELEMENTS_PER_CELL
assert reconstructed_data[start:end] == coset_evals
2024-04-19 11:50:22 +00:00
reconstructed_data_as_cells = [
coset_evals_to_cell(reconstructed_data[i * FIELD_ELEMENTS_PER_CELL:(i + 1) * FIELD_ELEMENTS_PER_CELL])
2024-04-19 11:50:22 +00:00
for i in range(CELLS_PER_EXT_BLOB)]
2024-04-19 11:50:22 +00:00
return reconstructed_data_as_cells
2023-12-12 16:25:30 +00:00
```