PedersenCommitment<T>

Models & Types · AiDotNet.FederatedLearning.Verification

Implements Pedersen commitment scheme — additively homomorphic for verifiable aggregation.

Inheritance: FederatedLearningComponentBase<T> → PedersenCommitment<T>

Implements: IGradientCommitment<T>, IZkProofSystem

For Beginners

Pedersen commitments are special because they’re “additively homomorphic”: if you commit to values a and b separately, anyone can combine the two commitments to get a valid commitment to a+b — without knowing a or b.

How It Works

How it works:

In FL: The server receives commitments from all clients. It can multiply them together to get a commitment to the sum. Then when clients open their commitments, the server verifies both individual values and the aggregate — detecting any manipulation.

Reference: Pedersen (CRYPTO 1991). Used in RiseFL (VLDB 2024) for scalable FL verification.

Constructors

Constructor Summary
PedersenCommitment(int groupBitLength) Initializes a new instance of PedersenCommitment.

Properties

Property Summary
string Name { get; }

Methods

Method Summary
GradientCommitmentData<T> Commit(Tensor<T> gradient)
Tensor<T> Open(GradientCommitmentData<T> commitment)
ValueTuple<byte[], byte[]> Commit(byte[] value)
bool Verify(GradientCommitmentData<T> commitment)
bool VerifyAggregation(IReadOnlyList<GradientCommitmentData<T>> individualCommitments, GradientCommitmentData<T> aggregateCommitment)
bool VerifyOpening(byte[] commitment, byte[] value, byte[] randomness)
bool VerifyRangeProof(byte[] proof, byte[] upperBound, byte[] commitment)
byte[] GenerateRangeProof(byte[] value, byte[] upperBound, byte[] commitment, byte[] randomness)