datasynth-banking

KYC/AML banking transaction generator for synthetic data.

Overview

datasynth-banking provides comprehensive banking transaction simulation for:

• Compliance testing and model training
• AML/fraud detection system evaluation
• KYC process simulation
• Regulatory reporting testing

Features

Architecture

BankingOrchestrator (orchestration)
        |
Generators (customer, account, transaction, counterparty)
        |
Typologies (AML pattern injection)
        |
Labels (ground truth generation)
        |
Models (customer, account, transaction, KYC)

Module Structure

Models

KYC Profile

#![allow(unused)]
fn main() {
pub struct KycProfile {
    pub declared_purpose: String,
    pub turnover_band: TurnoverBand,
    pub transaction_frequency: TransactionFrequency,
    pub expected_categories: Vec<TransactionCategory>,
    pub source_of_funds: SourceOfFunds,
    pub source_of_wealth: SourceOfWealth,
    pub geographic_exposure: Vec<String>,
    pub cash_intensity: CashIntensity,
    pub beneficial_owner_complexity: OwnerComplexity,
    // Ground truth fields
    pub is_deceiving: bool,
    pub actual_turnover_band: Option<TurnoverBand>,
}
}

AML Typologies

Labels

Usage Examples

Basic Generation

#![allow(unused)]
fn main() {
use synth_banking::{BankingOrchestrator, BankingConfig};

let config = BankingConfig::default();
let mut orchestrator = BankingOrchestrator::new(config, 12345);

// Generate customers and accounts
let customers = orchestrator.generate_customers();
let accounts = orchestrator.generate_accounts(&customers);

// Generate transaction stream
let transactions = orchestrator.generate_transactions(&accounts);
}

With AML Typologies

#![allow(unused)]
fn main() {
use synth_banking::{BankingConfig, TypologyConfig};

let config = BankingConfig {
    customer_count: 1000,
    typologies: TypologyConfig {
        structuring_rate: 0.02,   // 2% structuring patterns
        funnel_rate: 0.01,        // 1% funnel accounts
        mule_rate: 0.005,         // 0.5% mule networks
        ..Default::default()
    },
    ..Default::default()
};
}

Accessing Labels

#![allow(unused)]
fn main() {
let labels = orchestrator.generate_labels();

// Entity-level labels
for entity_label in &labels.entity_labels {
    println!("Customer {} risk: {:?}",
        entity_label.customer_id,
        entity_label.risk_tier
    );
}

// Transaction-level labels
for tx_label in &labels.transaction_labels {
    if tx_label.is_suspicious {
        println!("Suspicious: {} - {:?}",
            tx_label.transaction_id,
            tx_label.typology
        );
    }
}
}

Key Types

Customer Types

#![allow(unused)]
fn main() {
pub enum BankingCustomerType {
    Retail,     // Individual customers
    Business,   // Business accounts
    Trust,      // Trust/corporate entities
}
}

Risk Tiers

#![allow(unused)]
fn main() {
pub enum RiskTier {
    Low,
    Medium,
    High,
    Prohibited,
}
}

Transaction Categories

#![allow(unused)]
fn main() {
pub enum TransactionCategory {
    SalaryWages,
    BusinessPayment,
    Investment,
    RealEstate,
    Gambling,
    Cryptocurrency,
    CashDeposit,
    CashWithdrawal,
    WireTransfer,
    AtmWithdrawal,
    PosPayment,
    OnlinePayment,
    // ... more categories
}
}

AML Typologies

#![allow(unused)]
fn main() {
pub enum AmlTypology {
    Structuring,
    Funnel,
    Layering,
    Mule,
    RoundTripping,
    CreditCardFraud,
    SyntheticIdentity,
    None,
}
}

Export Files

See Also

• datasynth-core - Core banking models
• Fraud Detection Use Case
• Anomaly Injection