Open Source · Apache 2.0 · Federated Graph Protocol

Federated Risk Typology & Signal Registry

The open-source registry for federated financial fraud and AML typologies and detection signals. Version, distribute, and validate production-grade models across distributed financial institutions and digital asset networks.

Get started free Browse signals & models

Typologies

Models

1,259

Pulls

Nodes

Browse Signals

Fraud typologies and detection patterns across the federated network.

19 typologies

Signal	Slug	Upvotes	Pulls	Submissions
Cross-Border Wire Stacking Rapid sequential wire transfers across multiple jurisdictions to outpace investigative response	wire_stacking	6	21	33
Crypto Mixing / Tumbling Obfuscating crypto transaction trails through mixing services	crypto_mixing	44	31	65
Cuckoo Smurfing Exploiting legitimate remittance channels to co-mingle illicit funds	cuckoo_smurfing	17	21	62
Cycle Detection Circular transaction flows indicating coordinated laundering rings	cycle	40	34	66
Dormant Account Activation Long-inactive accounts suddenly receiving high-volume transactions	dormant_activation	45	50	72
Fan-in (Mule Collector) Many accounts funneling funds into a single mule account	fan_in_mule	33	30	63
Fan-out (Burst) Single account rapidly dispersing funds to many recipients	fan_out_burst	44	58	64
Funnel Account Single account aggregating deposits from dispersed geographic locations	funnel_account	35	27	65
Invoice Factoring Fraud Fraudulent or duplicate invoices used to obtain financing against fictitious receivables	invoice_factoring_fraud	6	15	25
Layering Chain Sequential hops through intermediary accounts to obscure fund origin	layering_chain	27	36	62
Mule Recruitment Network Social-engineered recruitment rings coordinating multiple money mules	mule_recruitment_network	2	15	22
Nesting / Payable-Through Respondent bank clients accessing correspondent banking indirectly	nesting	21	22	59
Peel Chain Large crypto amount peeled off in small increments across wallets	peel_chain	28	31	62
Rapid Movement Funds moved through multiple accounts within hours to avoid detection	rapid_movement	15	44	57
Round-Tripping Funds sent abroad and returned disguised as foreign investment	round_tripping	47	23	57
Shell Company Network Funds routed through entities with no real business operations	shell_company	32	95	62
Smurfing (Structuring) Coordinated small transactions structured below reporting thresholds	smurfing	39	31	51
Synthetic Identity Fraud Fabricated identities combining real and fake PII to open and operate accounts	synthetic_identity	3	13	21
Trade-Based Laundering Over/under-invoicing of goods and services to move value across borders	trade_based	33	21	51

Browse Models

Production-grade models registered across the federated graph.

11 models

fgi-autoencoder-anomaly

Production

Variational autoencoder for anomaly detection via reconstruction error on transaction embeddings

Autoencoder pytorch v3

50 pulls 120

View

fgi-gat-multihead

Production

Multi-head attention GAT with 8 attention heads for heterogeneous transaction graphs

GAT pytorch v3

57 pulls 104

View

fgi-gcn-spectral

Production

Spectral GCN with Chebyshev polynomial approximation for transaction graph convolution

GCN tensorflow v4

53 pulls 119

View

fgi-graphsage-inductive

Production

GraphSAGE inductive model for unseen node classification in expanding transaction graphs

GraphSAGE pytorch v3

75 pulls 109

View

fgi-hetero-gnn

Production

Heterogeneous graph neural network modeling distinct node/edge types across account, merchant, and device layers

hetero_gnn pytorch v1

27 pulls 31

View

fgi-onnx-edge

Production

ONNX-exported lightweight GNN for edge/on-device inference at partner institutions

GNN onnx v1

22 pulls 33

View

fgi-rf-ensemble

Production

Random forest ensemble on engineered graph/behavioral features — baseline benchmark model

ensemble sklearn v1

20 pulls 32

View

fgi-risk-gnn

Production

Federated Graph Intelligence risk scoring GNN — 128-dim node embeddings with 2-layer message passing

GNN pytorch v3

113 pulls 132

View

fgi-temporal-gat

Production

Temporal Graph Attention Network — time-aware attention over transaction sequences

Temporal GNN pytorch v5

80 pulls 110

View

fgi-txn-transformer

Production

Transaction-level Transformer encoder for sequential fraud pattern detection

Transformer pytorch v5

78 pulls 116

View

fgi-xgb-features

Production

XGBoost classifier on graph-derived feature vectors — degree centrality, PageRank, clustering coefficient

XGBoost xgboost v5

66 pulls 113

View

Browse Nodes

Regulatory jurisdictions participating in the federated graph network.

7 regions

US / FinCEN

us_fincen

United States — Financial Crimes Enforcement Network jurisdiction

EU / AMLD

eu_amld

European Union — Anti-Money Laundering Directive jurisdiction

UK / FCA

uk_fca

United Kingdom — Financial Conduct Authority jurisdiction

APAC / FATF

apac_fatf

Asia-Pacific — FATF-aligned regulatory framework

LATAM

latam

Latin America — Regional AML/CFT framework

MENA

mena

Middle East & North Africa — Regional regulatory framework

Africa

africa

Sub-Saharan Africa — Regional AML/CFT framework

Infrastructure for Federated Payments Fraud Intelligence

Everything you need to version, validate, and distribute typologies across the federated graph.

Federated Aggregation

Coordinate typology updates across distributed graph nodes with cryptographically versioned weights.

Immutable Signal Registry

SHA-verified, append-only storage for typology artifacts with granular RBAC and audit trails.

Edge-Ready Distribution

Hot-swap production typologies at the edge. Sub-second propagation across all participating nodes.

Continuous Validation

Automated drift detection and quality gates on every submission before promotion to production.

Quickstart

Get up and running with the official federated-graph Python library.

Install the library

pip install federated-graph

Optional extras: [grpc] [redis] [scoring] [all]

Configure and initialize a model

import torch
from torch_geometric.data import Data
from federated_graph import FGIConfig
from federated_graph.models import FGIModel

# Create configuration
config = FGIConfig()

# Initialize model
model = FGIModel(in_channels=10, config=config.model)

Run a forward pass on graph data

# Prepare graph data
x = torch.randn(100, 10)
edge_index = torch.randint(0, 100, (2, 500))

# Forward pass returns embeddings and risk scores
embeddings, entity_scores, txn_scores = model(x, edge_index)

Start federated training

from federated_graph.federation import FederationServer, FederationClient

server = FederationServer(
    num_rounds=config.federation.num_rounds,
    min_clients=config.federation.min_clients,
)

client = FederationClient(
    client_id="client_1",
    model=model,
    config=config,
)

Full documentation at pypi.org/project/federated-graph

Start contributing to the graph

Join the network. Push typologies, pull signals, validate updates — all from your terminal.

Create free account