AdaptiveCompressor<T> |
Adaptive compressor: dynamically adjusts compression ratio per client based on bandwidth, gradient importance, and staleness. |
AesCtrSecureAggregationPrg |
AES-256 in counter mode as the secure-aggregation pseudorandom generator. |
AgnosticFairnessObjective<T> |
Implements AFL (Agnostic Federated Learning) — minimax fairness optimization. |
AmdSevSnpTeeProvider<T> |
TEE provider for AMD SEV-SNP (Secure Encrypted Virtualization - Secure Nested Paging). |
ArithmeticSecretSharing<T> |
Implements additive secret sharing over an arithmetic field for efficient linear operations. |
ArmCcaTeeProvider<T> |
TEE provider for ARM CCA (Confidential Compute Architecture) / Realm Management Extension. |
AsyncFedEDTrainer<T> |
Implements AsyncFedED — Asynchronous FL with Entropy-Driven client scheduling. |
AvailabilityAwareClientSelectionStrategy |
Availability-aware client selection using per-client online probabilities. |
BaseObliviousTransfer |
Implements base 1-out-of-2 oblivious transfer using symmetric cryptography. |
BasicCompositionPrivacyAccountant |
Privacy accountant using basic (naive) composition. |
BloomFilterPsi |
Implements Bloom filter based probabilistic Private Set Intersection. |
BobaAggregationStrategy<T> |
Implements BOBA (Bayesian Optimal Byzantine-robust Aggregation) strategy. |
BooleanSecretSharing |
Implements XOR-based boolean secret sharing for bitwise operations. |
BooleanTriple |
Represents a pre-shared AND triple for boolean secret sharing. |
BucketingAggregationStrategy<T> |
Implements the Bucketing meta-strategy for Byzantine-robust federated learning. |
BufferedAsyncFederatedTrainer<T> |
Implements FedBuff — Buffered asynchronous federated aggregation. |
BulyanFullModelAggregationStrategy<T, TInput, TOutput> |
Bulyan aggregation for IFullModel (Multi-Krum selection + closest-to-median averaging). |
CircuitBasedPsi |
Implements circuit-based Private Set Intersection using garbled circuit evaluation. |
CircuitGate |
Represents a single gate in a boolean circuit. |
ClientDriftResult |
Per-client drift analysis result. |
ClientGraphStats |
Statistics about a client’s local subgraph, used for graph-aware aggregation weighting. |
ClientProofBundle |
Contains all proofs that a client must provide for a given verification level. |
ClientVerificationResult |
Contains the result of verifying a single client’s update. |
ClusteredClientSelectionStrategy |
Cluster-based client selection using simple k-means over per-client embeddings. |
CompressedTree<T> |
Represents a compressed decision-tree encoding of a parameter update. |
ComputationIntegrityProof<T> |
Generates proofs of local training computation integrity (research-stage). |
ContributionBasedIncentive<T> |
Incentive mechanism that rewards clients proportional to their evaluated contribution. |
DFedAvgMProtocol<T> |
Implements DFedAvgM — Decentralized FedAvg with Momentum for peer-to-peer FL. |
DPFedLoRA<T> |
Implements DP-FedLoRA — Differentially Private Federated LoRA aggregation. |
DataFreeFCL<T> |
Implements Data-Free Federated Continual Learning — prevents forgetting without storing real data. |
DataShapleyEvaluator<T> |
Data Shapley evaluator: efficient Monte Carlo approximation of Shapley values. |
DeTAGProtocol<T> |
Implements DeTAG — Decentralized gradient Tracking for exact convergence. |
DecentralizedAggregator<T> |
Decentralized aggregator — performs local mixing of model parameters based on topology. |
DiffieHellmanPsi |
Implements Diffie-Hellman based Private Set Intersection using commutative encryption. |
DiffusiveNoiseUnlearner<T> |
Unlearning via structured diffusive noise injection targeting memorized samples. |
DirectionAlignmentInspector<T> |
Direction Alignment Inspector — detects backdoor attacks via gradient direction analysis. |
DivideAndConquerAggregationStrategy<T> |
Implements DnC (Divide and Conquer) aggregation strategy for Byzantine-robust FL. |
DriftAdaptiveAggregator<T> |
Drift-adaptive aggregator: wraps any aggregation strategy and adjusts weights based on drift. |
DriftReport |
Comprehensive report of drift analysis across all federated clients. |
EntityAligner |
High-level orchestrator for entity alignment in vertical federated learning. |
EntityAlignmentResult |
Contains the results of an entity alignment operation, including the PSI result, party sizes, and diagnostic information. |
ErrorFeedbackCompressor<T> |
Error feedback compressor: wraps any compression method with residual accumulation. |
ExactRetrainingUnlearner<T> |
Gold-standard unlearning: retrains the model from scratch excluding the target client. |
ExpandedSubgraph<T> |
Represents a subgraph expanded with pseudo-nodes. |
ExtendedObliviousTransfer |
Implements OT extension — amortizes a small number of base OTs into many cheap OTs. |
FLTrustAggregationStrategy<T> |
Implements the FLTrust aggregation strategy for Byzantine-robust federated learning. |
FLoRA<T> |
Implements FLoRA — Federated Low-Rank Adaptation with stacked lossless aggregation. |
FedAGCContinualLearning<T> |
Implements FedAGC — Adaptive Gradient Correction for federated continual learning. |
FedAGHNPersonalization<T> |
Implements FedAGHN (Adaptive Gradient-based Heterogeneous Networks) personalization. |
FedAaAggregationStrategy<T> |
Implements FedAA (Federated Adaptive Aggregation) strategy. |
FedAdagradServerOptimizer<T> |
FedAdagrad server optimizer — adaptive learning rates using accumulated squared gradients. |
FedAdamServerOptimizer<T> |
FedAdam server optimizer — adaptive learning rates with momentum and second-moment estimation. |
FedAlignAggregationStrategy<T> |
Implements FedAlign (Feature Alignment) aggregation strategy. |
FedAvgAggregationStrategy<T> |
Implements the Federated Averaging (FedAvg) aggregation strategy. |
FedAvgFullModelAggregationStrategy<T, TInput, TOutput> |
FedAvg aggregation for IFullModel using vector-based parameters. |
FedAvgMServerOptimizer<T> |
FedAvgM server optimizer — server-side momentum for stabilized federated averaging. |
FedBABUPersonalization<T> |
Implements FedBABU (Body And Bottom Update) personalization strategy. |
FedBNAggregationStrategy<T> |
Implements the Federated Batch Normalization (FedBN) aggregation strategy. |
FedBNFullModelAggregationStrategy<T, TInput, TOutput> |
FedBN aggregation for IFullModel when the model is a NeuralNetworkBase. |
FedCILContinualLearning<T> |
Implements FedCIL — federated class-incremental learning with generative replay, model consolidation and consistency enforcement. |
FedCILGenerativeModel<T> |
An auxiliary-classifier GAN used by FedCILContinualLearning: a conditional generator G(y, ε) and a discriminator with a shared hidden layer feeding a real/fake head and a class head. |
FedCPPersonalization<T> |
Implements FedCP (Conditional Policy) personalization with input-dependent routing. |
FedDFDistillation<T> |
FedDF — Federated ensemble distillation using model averaging on unlabeled public data. |
FedDTCompressor<T> |
Implements FedDT — Decision-tree-based compression for heterogeneous federated architectures. |
FedDynHeterogeneityCorrection<T> |
FedDyn-style dynamic regularization using a per-client drift accumulator. |
FedFairOptimizer<T> |
Implements FedFair — multi-objective optimization balancing accuracy, fairness, and efficiency. |
FedGENDistillation<T> |
FedGEN — Data-free federated distillation using a lightweight generator on the server. |
FedGnnAggregationStrategy<T> |
GNN-aware federated aggregation that weights contributions by subgraph topology characteristics. |
FedKDCompressedGradient<T> |
A gradient matrix compressed by FedKD’s truncated SVD: g ≈ U diag(S) V. |
FedKDCompressor<T> |
Implements FedKD — communication-efficient federated learning via adaptive mutual knowledge distillation and dynamic SVD gradient compression. |
FedKDLosses<T> |
The losses of one FedKD adaptive mutual distillation step, averaged over the batch. |
FedLcAggregationStrategy<T> |
Implements FedLC (Logit Calibration) aggregation strategy. |
FedMDDistillation<T> |
FedMD — Model-agnostic federated learning via mutual distillation on a public dataset. |
FedMaAggregationStrategy<T> |
Implements FedMA (Federated Matched Averaging) aggregation strategy. |
FedMeZO<T> |
Implements FedMeZO — Memory-efficient Zeroth-Order optimization for federated LLM fine-tuning. |
FedNovaHeterogeneityCorrection<T> |
FedNova-style normalization of client updates by local steps. |
FedPACPersonalization<T> |
Implements FedPAC (Personalization via Aggregation and Calibration) with prototype alignment. |
FedPETuning<T> |
Implements FedPETuning — a unified framework for parameter-efficient fine-tuning (PEFT) in federated learning that supports multiple PEFT methods under one API. |
FedProxAggregationStrategy<T> |
Implements the Federated Proximal (FedProx) aggregation strategy. |
FedProxFullModelAggregationStrategy<T, TInput, TOutput> |
FedProx aggregation for IFullModel. |
FedRoDPersonalization<T> |
Implements FedRoD (Representation on Demand) personalization with dual classifiers. |
FedSamAggregationStrategy<T> |
Implements FedSAM (Sharpness-Aware Minimization for Federated Learning) aggregation strategy. |
FedSelectPersonalization<T> |
Implements FedSelect — learned sparse binary masks for personalization. |
FedYogiServerOptimizer<T> |
FedYogi server optimizer — adaptive learning rates with controlled second-moment growth. |
FederatedAdapterTuning<T> |
Implements FedAdapter — federated bottleneck adapter tuning where small adapter modules are inserted into each transformer block and only these are communicated. |
FederatedAdvancedCapabilities<T> |
Drives the advanced federated-learning capabilities (contribution, fairness, drift, PSI, TEE, ZK, MPC, unlearning) at their real points in the federated training lifecycle, on behalf of the trainer, and collects their results into a `Federa… |
FederatedAdvancedMetadata |
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 c… |
FederatedDPO<T> |
Implements Federated DPO (Direct Preference Optimization) for reward-model-free LLM alignment. |
FederatedEWC<T> |
Federated Elastic Weight Consolidation (EWC) — prevents forgetting by penalizing changes to parameters that are important for previously learned tasks. |
FederatedExperienceReplay<T> |
Implements Federated Experience Replay for continual learning. |
FederatedGraphPartitioner<T> |
Partitions a graph across federated clients using various strategies. |
FederatedLoRA<T> |
Federated LoRA — Low-Rank Adaptation for parameter-efficient federated fine-tuning. |
FederatedOrthogonalProjection<T> |
Federated Orthogonal Projection — prevents forgetting by projecting gradients to be orthogonal to the subspace of previously important parameter directions. |
FederatedPromptTuning<T> |
Federated Prompt Tuning — soft prompt aggregation for foundation model personalization. |
FederatedRLHF<T> |
Configuration and orchestration for Federated RLHF (Reinforcement Learning from Human Feedback). |
FetchSGDCompressor<T> |
Implements FetchSGD — Count-Sketch compression with momentum and error accumulation in sketch space. |
FixedPointRingEncoding |
Fixed-point encoding of real-valued model updates into the ring Z_R, R = 2^b, in which Bonawitz et al. |
FlameAggregationStrategy<T> |
Implements FLAME (Filtering via cosine similarity + Adaptive clipping + Noise) for Byzantine-robust federated learning with backdoor resistance. |
FuzzyPsi |
Implements approximate entity matching using multiple similarity strategies. |
GarbledCircuitData |
Contains the garbled circuit data produced by the garbler. |
GarbledCircuitEvaluator |
Evaluates garbled circuits produced by GarbledCircuitGenerator. |
GarbledCircuitGenerator |
Implements Yao’s garbled circuit generation with point-and-permute, free XOR, and half-gates optimizations. |
GaussianDifferentialPrivacyVector<T> |
Implements Gaussian differential privacy for vector-based model updates. |
GaussianDifferentialPrivacy<T> |
Implements differential privacy using the Gaussian mechanism. |
GossipProtocol |
Gossip Protocol — randomized peer-to-peer model exchange for decentralized FL. |
GradientAscentUnlearner<T> |
Approximate unlearning via gradient ascent: reverses learning by ascending the loss on target data. |
GradientBoundednessProof<T> |
Proves that each gradient component is within [-B, B] (element-wise range proof). |
GradientCommitmentData<T> |
Contains the data for a gradient commitment. |
GradientNormRangeProof<T> |
Generates and verifies proofs that a gradient’s L2 norm is within a declared bound. |
GradientSketchCompressor<T> |
Count Sketch-based gradient compression for federated learning. |
GraphNeighborhoodPrivacy<T> |
Applies local differential privacy (LDP) to neighborhood queries to prevent topology leakage. |
GraphNodeGenerator<T> |
Generates pseudo-node features for missing cross-client neighbors using a learned model. |
GroupFairnessConstraint<T> |
Enforces group fairness constraints during federated learning aggregation. |
HashCommitmentScheme<T> |
Implements a hash-based commitment scheme using SHA-256. |
HeterogeneousLoRA<T> |
Heterogeneous LoRA — supports different LoRA ranks per client with SVD-based aggregation. |
HierarchicalFedLoRA<T> |
Implements HierFedLoRA — Hierarchical LoRA aggregation for edge-cloud federated topologies. |
HybridMpcProtocol<T> |
Combines arithmetic secret sharing (for linear operations) with garbled circuits (for non-linear operations) into a single hybrid MPC protocol. |
InMemoryFederatedTrainer<T, TInput, TOutput> |
In-memory federated learning trainer for local simulation and tests. |
InfluenceFunctionUnlearner<T> |
Influence function-based unlearning: mathematically estimates and subtracts a client’s contribution. |
IntelSgxTeeProvider<T> |
TEE provider for Intel SGX (Software Guard Extensions) process-level enclaves. |
IntelTdxTeeProvider<T> |
TEE provider for Intel TDX (Trust Domain Extensions) confidential VMs. |
KNNPersonalization<T> |
Implements kNN-Per — kNN-based personalization at inference time with zero extra training cost. |
KrumFullModelAggregationStrategy<T, TInput, TOutput> |
Krum aggregation for IFullModel (Byzantine-robust selection by distance). |
LabelDifferentialPrivacy<T> |
Implements differential privacy protection for label holder gradients in VFL. |
LeafFederatedDatasetLoader<T> |
Loads LEAF benchmark JSON files into per-client datasets suitable for federated learning simulation. |
LeafFederatedDataset<TInput, TOutput> |
Represents a LEAF dataset with optional train/test splits. |
LeafFederatedSplit<TInput, TOutput> |
Represents a single LEAF split (train/test) as per-client datasets. |
LeafRedditFederatedDatasetLoader |
Loads the LEAF Reddit benchmark JSON files into per-client token-sequence datasets. |
LeafSent140FederatedDatasetLoader |
Loads the LEAF Sent140 benchmark JSON files into per-client datasets. |
LeafShakespeareFederatedDatasetLoader |
Loads the LEAF Shakespeare benchmark JSON files into per-client datasets. |
LeafTokenSequenceFederatedDatasetLoader |
Loads LEAF-style JSON files that store token sequences (x) and next-token labels (y). |
LightweightShapleyEvaluator<T> |
Implements Lightweight Shapley — O(n) Shapley value approximation using gradient similarity. |
LossThresholdProof<T> |
Proves that a client’s local training loss is below a threshold without revealing the actual value. |
MedianFullModelAggregationStrategy<T, TInput, TOutput> |
Coordinate-wise median aggregation for IFullModel. |
MissingFeatureHandler<T> |
Handles missing features in vertical FL when not all parties have data for all entities. |
ModelDriftDetector<T> |
Model-based drift detector: uses gradient direction and weight divergence to detect drift. |
ModelUpdateVerifier<T> |
Server-side verification engine that checks all proofs from clients before aggregation. |
MoonAggregationStrategy<T> |
Implements MOON (Model-COntrastive Learning) aggregation strategy. |
MultiKrumFullModelAggregationStrategy<T, TInput, TOutput> |
Multi-Krum aggregation for IFullModel (select m central updates, then average). |
MultiPartyPsi |
Implements multi-party Private Set Intersection for 3 or more parties. |
MultiPartyPsiResult |
Contains results of a multi-party PSI computation. |
NeuralCleanseDetector<T> |
Neural Cleanse — post-hoc backdoor detection by reverse-engineering potential triggers. |
ObliviousTransferPsi |
Implements Oblivious Transfer based Private Set Intersection using cuckoo hashing. |
OpenFedLLMPipeline<T> |
Implements OpenFedLLM pipeline patterns for federated LLM training, alignment, and serving. |
OptiGradTrustAggregationStrategy<T> |
Implements OptiGradTrust (Optimized Gradient Trust) aggregation strategy with historical reputation tracking. |
OptimizedPrivateSetAnalytics<T> |
Implements Optimized Private Set Analytics (OPSA) beyond basic intersection. |
PFedGatePersonalization<T> |
Implements pFedGate — gated layer-wise mixture of local and global parameters. |
PedersenCommitment<T> |
Implements Pedersen commitment scheme — additively homomorphic for verifiable aggregation. |
PerformanceAwareClientSelectionStrategy |
Performance-aware client selection using an explore/exploit policy over historical scores. |
PersonalizedFederatedLearning<T> |
Implements personalized federated learning where each client maintains some client-specific parameters. |
PowerSGDCompressor<T> |
PowerSGD: low-rank gradient compression using randomized SVD approximation. |
PrototypeFederatedGraphLearning<T> |
Prototype-based federated graph learning: clients share class prototypes instead of full model parameters. |
PrototypicalContributionEvaluator<T> |
Prototypical contribution evaluator: measures client value using prototype representations. |
ProxyCertificate |
A signed certificate issued by the proxy after verifying a client update. |
ProxyVerificationResult |
Result of a proxy ZKP verification. |
ProxyZKPVerifier<T> |
Implements Proxy-based Zero-Knowledge Proof verification for federated learning. |
PsiResult |
Contains the results of a Private Set Intersection computation. |
QFairFederatedLearning<T> |
Implements q-FFL (q-Fair Federated Learning) — parameterized fairness via power-mean. |
RdpPrivacyAccountant |
A simple Rényi Differential Privacy (RDP) accountant for repeated Gaussian mechanisms. |
RemoteAttestationResult |
Represents the result of verifying a remote attestation quote from a TEE. |
RfaFullModelAggregationStrategy<T, TInput, TOutput> |
Robust Federated Aggregation (RFA) via geometric median (Weiszfeld iterations). |
RingAllReduceProtocol |
Ring AllReduce — communication-efficient decentralized averaging using a ring topology. |
SampledSubgraph<T> |
Represents a sampled subgraph from neighborhood sampling. |
ScaffoldHeterogeneityCorrection<T> |
SCAFFOLD-style heterogeneity correction using control variates. |
SecureAggregationAdvertisedKeys |
Round 0 (AdvertiseKeys) message: a client’s two P-256 public keys. |
SecureAggregationEncryptedShare |
Round 1 (ShareKeys) message: client u’s shares for client v, encrypted so only v can read them. |
SecureAggregationMaskedInput |
Round 2 (MaskedInputCollection) message: a client’s masked input y_u, a vector in Z_R^m. |
SecureAggregationUnmaskingResponse |
Round 4 (Unmasking) message: the shares a surviving client releases so the server can remove the masks. |
SecureAggregation<T> |
In-process secure aggregation following Bonawitz et al. |
SecureClippingProtocol<T> |
Implements secure gradient clipping without revealing gradient norms. |
SecureComparisonProtocol<T> |
Implements secure greater-than comparison on secret-shared values. |
SecureCrossClientEdgeDiscovery<T> |
PSI-based secure discovery of cross-client edges without revealing full adjacency lists. |
SecureGradientExchange<T> |
Provides encryption for gradient tensors exchanged between parties in vertical FL. |
SegmentedGossipProtocol<T> |
Implements Segmented Gossip — communication-efficient gossip that exchanges model segments. |
SemiAsyncFederatedTrainer<T> |
Implements Semi-Asynchronous Federated Learning — hybrid sync/async with periodic barriers. |
ShapleyValueEvaluator<T> |
Exact Shapley value evaluator: computes each client’s marginal contribution across all coalitions. |
ShuffleModelDP<T> |
Implements Shuffle Model Differential Privacy for federated learning. |
SimulatedTeeProvider<T> |
Software-simulated TEE provider for testing and development without hardware. |
SparseLoRA<T> |
Implements SLoRA — Sparse LoRA for communication-efficient federated fine-tuning. |
SparseUpdate<T> |
Represents a sparse update: only a subset of indices have non-zero values. |
SplitNeuralNetwork<T> |
Implements a split neural network for vertical federated learning. |
StatisticalDriftDetector<T> |
Statistical drift detector: uses Page-Hinkley, ADWIN, or DDM tests on client metrics. |
StratifiedClientSelectionStrategy |
Stratified client selection using a client-to-group mapping. |
SubgraphExpander<T> |
Expands a local subgraph with pseudo-nodes to approximate missing cross-client neighbors. |
SubgraphFederatedTrainer<T> |
Main coordinator for subgraph-level federated GNN training. |
TeeSecureAggregation<T> |
Performs weighted model aggregation inside a TEE enclave boundary. |
TiltedERMFairness<T> |
Implements TERM (Tilted Empirical Risk Minimization) for fairness in FL. |
TimeVaryingTopology<T> |
Implements time-varying topology for decentralized federated learning. |
TopKSparsificationCompressor<T> |
Implements Top-k Sparsification — send only the k largest gradient elements. |
TrainingStep |
Represents a single training step in the integrity log. |
TrainingStepLog |
Records training steps for computation integrity verification. |
TrimmedMeanFullModelAggregationStrategy<T, TInput, TOutput> |
Coordinate-wise trimmed mean aggregation for IFullModel. |
UnbalancedPsi |
Implements PSI optimized for asymmetric (unbalanced) set sizes. |
UniformRandomClientSelectionStrategy |
Uniform random client selection (fractional participation). |
UnlearningCertificate |
Certificate proving that a client’s data has been unlearned from the federated model. |
UnlearningVerification |
Contains the results of an unlearning verification check. |
VerifiableAggregationStrategy<TModel> |
Decorator that wraps any IAggregationStrategy with proof verification. |
VerificationConstraint |
Describes a constraint to be proven in zero-knowledge. |
VerificationProof |
Represents a zero-knowledge proof of a computational property. |
VerificationRecord |
Records a verification event for auditing. |
VerticalDataPartitioner<T> |
Partitions features (columns) across parties for vertical federated learning simulation. |
VerticalFederatedBenchmark<T> |
Provides benchmarking utilities for evaluating VFL implementations. |
VerticalFederatedTrainer<T> |
Main orchestrator for vertical federated learning training. |
VerticalFederatedUnlearner<T> |
Implements GDPR-compliant entity unlearning for vertical federated learning models. |
VerticalPartyClient<T> |
Represents a feature-holding party in vertical federated learning. |
VerticalPartyLabelHolder<T> |
Represents the label-holding party in vertical federated learning. |
VflAlignmentSummary |
Contains summary statistics from the entity alignment phase of VFL. |
VflBenchmarkDataset<T> |
Contains a complete benchmark dataset with vertically-partitioned data. |
VflBenchmarkResult |
Contains results from a VFL benchmark run. |
VflEpochResult<T> |
Contains metrics from a single VFL training epoch. |
VflPartyDataset<T> |
Contains data for a single party in a benchmark dataset. |
VflTrainingResult<T> |
Contains the complete results from a VFL training run. |
WeightedRandomClientSelectionStrategy |
Weighted random selection without replacement (typically weighted by sample count). |
WinsorizedMeanFullModelAggregationStrategy<T, TInput, TOutput> |
Coordinate-wise winsorized mean aggregation for IFullModel. |
ZOClientMessage |
Compact message from a client in the FedMeZO protocol. |