VerifiableAggregationStrategy<TModel>

Models & Types · AiDotNet.FederatedLearning.Verification

Decorator that wraps any IAggregationStrategy with proof verification.

Implements: IAggregationStrategy<TModel>

For Beginners

This class wraps an existing aggregation strategy (like FedAvg or Krum) and adds verification: before aggregating updates, it checks cryptographic proofs from each client. Only clients that pass verification are included in the aggregation.

How It Works

How to use:

Integration with existing FL: The existing Byzantine-robust aggregators (Krum, Bulyan, Median) detect statistical anomalies. This adds cryptographic guarantees on top — first verify proofs, then pass verified updates to the inner aggregator.

Constructors

Constructor Summary
VerifiableAggregationStrategy(IAggregationStrategy<TModel> innerStrategy, VerificationOptions options) Initializes a new instance of VerifiableAggregationStrategy.

Properties

Property Summary
IReadOnlyList<int> RejectedClients { get; } Gets the list of client IDs rejected in the current round.
VerificationOptions Options { get; } Gets the verification options.

Methods

Method Summary
TModel Aggregate(Dictionary<int, TModel> clientModels, Dictionary<int, double> clientWeights)
string GetStrategyName()
void RegisterClientProof(int clientId, ClientProofBundle proofBundle) Registers a client’s proof bundle for verification during aggregation.
void SetRound(int round) Sets the current training round (for proof verification context).