AMLNet - Synthetic Anti-Money Laundering Transaction Dataset

**DEPRECATED (Superseded): This record (mislabelled as Version 1.0) is superseded by the corrected release: DOI https://doi.org/10.5281/zenodo.16736515. Do not use this deprecated version for analysis; it contains incomplete configuration/content.

AMLNet - Synthetic Anti-Money Laundering Transaction Dataset

DESCRIPTION:
This dataset contains over 1 million synthetic financial transactions (1,048,575) generated using the AMLNet framework for anti-money laundering research.

CONTENTS:
- 1,047,028 legitimate transactions across 5 categories
- 1,547 labeled money laundering transactions (0.15%)
- 192-day simulation period
- AUSTRAC-compliant suspicious patterns

DATA FORMAT:
CSV file with 16 columns containing transaction details:

CORE TRANSACTION DATA:
- step: Sequential transaction step/ID
- type: Payment method (TRANSFER, OSKO, BPAY, EFTPOS, DEBIT, NPP)
- amount: Transaction amount in AUD (positive/negative values)
- category: Transaction category (Housing, Food, Transport, Recreation, Other)
- nameOrig: Originating customer ID (e.g., C3511)
- nameDest: Destination customer/merchant ID (e.g., C4945, M558)
- oldbalanceOrg: Account balance before transaction
- newbalanceOrig: Account balance after transaction

LABELS:
- isFraud: Binary fraud indicator (0=legitimate, 1=fraudulent)
- isMoneyLaundering: Binary AML label (0=normal, 1=suspicious)
- fraud_probability: Calculated fraud risk score

TEMPORAL FEATURES:
- hour: Hour of transaction (0-23)
- day_of_week: Day of week (1=Monday, 7=Sunday)
- day_of_month: Day of month (1-31)
- month: Month number (1-12)

METADATA:
- metadata: JSON object containing:
  * timestamp: Exact transaction datetime
  * location: City, state, country, postcode
  * device_info: Device type, OS, IP address
  * payment_method: Specific payment method used
  * merchant_info: Merchant details (if applicable)
  * risk_indicators: Comprehensive risk scoring metrics

DATASET STATISTICS:
- Total transactions: 1,048,575 (1M+)
- Legitimate transactions: 1,047,028 (99.85%)
- Money laundering transactions: 1,547 (0.15%)
- CSV file rows: 1,048,576 (including header row)
- Payment types: 6 different methods
- Transaction categories: 5 main categories
- Time period: 192-day simulation
- Geographic coverage: Australian cities and postcodes

USAGE:
This dataset is designed for:
- Anti-money laundering research and algorithm development
- Financial fraud detection benchmarking
- Machine learning model training and validation
- Academic research in financial crime detection
- Commercial AML system development and testing

Licensed under CC BY 4.0. Free to use for any purpose with proper attribution.
See LICENSE.txt for full terms.

CITATION:
If you use this dataset, please cite:
Huda, S., Foo, E., Jadidi, Z., Newton, M.A.H., & Sattar, A. (2025). 
AMLNet: A Knowledge-Based Multi-Agent Framework to Generate and Detect 
Realistic Money Laundering Transactions. Expert Systems with Applications.

CONTACT:
s.huda@griffith.edu.au

VERSION: 1.0
DATE: July 2025