Models / Anti-Money Laundering

Mule Account Detector

Finds accounts behaving like money mules.

Strictly ConfidentialTier 1 modelFinancial crimeBanking secrecy+1High riskAnomaly Detection · 1.4.0 · mule-account-detector
38 96K/30d 51

Precision@top 200 daily

0.58

Recall of confirmed mules (90 days)

0.74

Alert volume vs rules

-46%

Median time to detection

2.1 days

About this model

Flags current and savings accounts that behave like money mule accounts: rapid pass-through of many small inbound DuitNow transfers, immediate cash-out, new accounts receiving from many unrelated senders. Combines account-level features with a graph of fund flows between accounts and returns a score, reason codes and the linked accounts behind the alert.

Intended use

Daily and intraday screening of accounts for Financial Crime investigators, and input to the Bank's scam reporting process. Investigators decide on freezes and reports; the model never freezes an account itself.

Training data lineage

Graph neural network plus gradient-boosted account features, trained in Vertex AI on casa-transactions fund flows in BigQuery, labelled with mule cases from confirmed-fraud-labels and closed investigations in aml-alerts-history. Tier 1 model validated by Model Risk Management; fairness review confirmed no reliance on age, gender or race proxies.

Limitations & bias notes

Mule accounts recruited recently with little history are hard to separate from new genuine customers; those are scored with lower confidence. Flows through e-wallets and other banks are only visible at the edge of the graph.

#mule-accounts#scams#graph#financial-crime#duitnow#reason-codes#tier-1

Ownership and sensitivity

Owned by

Group Financial Crime ComplianceAnti-Money Laundering

Accountable owner: Ravi Chandran

Who approves access

  1. Owner, Anti-Money Laundering, Ravi Chandran
  2. Data Governance Office, Ahmad Faizal
  3. Model Risk Management, Dr. Kavitha Subramaniam

Built on by Fraud Risk Management.

In Fight financial crime

Strictly Confidential

Customer-level data under banking secrecy and PDPA, and models that decide on customers.

Named-user entitlement with the data owner's consent and a stated purpose; reviewed on expiry.

Tier 1 model
Material model: independent validation before use and annual revalidation.
Financial crime
Used in AML, fraud or sanctions controls; tipping-off restrictions apply.
Banking secrecy
Customer account information protected under the Financial Services Act 2013.
Customer PII
Contains or processes personal data about customers (PDPA 2010).

Try it

Live sandbox

account=A-TKN-5521&age_days=42

fixture

Evaluation only: restricted models run against synthetic or masked sample data until your business unit's access request is approved.

Feedback

Details

Updated
2026-07-11
Latest version
1.4.0
Licence
BU Restricted
Access
Restricted
Framework
PyTorch
Language
N/A