FederatedAdvancedMetadata
Models & Types · AiDotNet.FederatedLearning
Results from the advanced federated-learning capabilities configured via
ConfigureFederatedLearning(...): client contribution, fairness, drift, private-set-intersection
overlap, TEE attestation, zero-knowledge proofs, secure multi-party computation, and unlearning readiness.
How It Works
Each block is populated only when the corresponding capability was configured; unconfigured capabilities
leave their *Enabled/*Evaluated flag false. This is surfaced on
FederatedLearningMetadata.
Constructors
| Constructor | Summary |
|---|---|
FederatedAdvancedMetadata() |
Properties
| Property | Summary |
|---|---|
Dictionary<int, double> ClientContributions { get; set; } |
Per-client contribution scores (higher = more valuable to the global model). |
HashSet<int> FreeRiders { get; set; } |
Clients identified as free-riders (contribution below the method’s threshold). |
bool AnyDriftDetected { get; set; } |
Whether drift was detected in any round. |
bool ContributionEvaluated { get; set; } |
Whether client contributions were evaluated. |
bool DriftDetectionEnabled { get; set; } |
Whether drift detection ran. |
bool FairnessEvaluated { get; set; } |
Whether a fairness constraint was evaluated. |
bool FairnessSatisfied { get; set; } |
Whether the final model satisfies the fairness constraint. |
bool McpEnabled { get; set; } |
Whether secure MPC aggregation was exercised. |
bool McpSecureSumVerified { get; set; } |
Whether the secure secret-shared sum reconstructed to the plaintext sum (correctness check). |
bool PsiEnabled { get; set; } |
Whether PSI ran. |
bool TeeEnabled { get; set; } |
Whether TEE attestation ran. |
bool UnlearningAvailable { get; set; } |
Whether an unlearner is configured (GDPR right-to-be-forgotten available on the result). |
bool ZkAllVerified { get; set; } |
Whether every produced proof verified. |
bool ZkEnabled { get; set; } |
Whether zero-knowledge proofs were produced. |
double FairnessViolation { get; set; } |
The fairness violation score (0 = perfectly fair, higher = more unfair). |
int DriftDetectedCount { get; set; } |
The number of rounds in which drift was detected. |
int DriftRoundsChecked { get; set; } |
The number of rounds drift was checked. |
int PsiTotalOverlap { get; set; } |
The total number of overlapping samples found across clients (data-leakage / dedup signal). |
int TeeAttestationCount { get; set; } |
The number of attestation quotes produced over the aggregated model. |
int ZkProofCount { get; set; } |
The number of range proofs produced (one per round, bounding the update norm). |
string ContributionMethod { get; set; } |
The contribution method’s name. |
string FairnessConstraint { get; set; } |
The fairness constraint’s name. |
string McpProtocol { get; set; } |
The MPC protocol’s name. |
string TeeMeasurementHash { get; set; } |
The enclave measurement (code-identity) hash. |
string UnlearningMethod { get; set; } |
The unlearning method’s name. |
string ZkSystem { get; set; } |
The zero-knowledge system’s name. |