4 Ways to Stop Money Mules

4 Ways Financial Institutions Can Stop Money Mules in Their Tracks

Financial institutions (FIs) face money mule and money laundering risks due to the real-time availability of their digital products and services.

The bad news: many anti-money laundering (AML) and mule detection solutions aren’t up to the task. As a result, most FIs are unable to immediately detect and stop incoming funds from illicit sources or customer accounts which are exhibiting money mule behavior.
The good news: there’s a better way forward. Here are 4 ways FIs can stop money mules in their tracks.

1: Use Real-Time Detection, Not Reactive Solutions

Traditional fraud prevention and AML methods are reactive and ineffective against money mules. These solutions identify and flag unusual long-term account activity associated with money laundering. With traditional methods, the goal is not to stop money laundering or muling as it happens, but rather to learn about the criminal network and share information with law enforcement for wider takedown efforts. This reactive approach allows mules to launder money unchecked, creating a vicious cycle of crime. FIs need a real-time approach to the real-time money mule problem. This is only possible with advanced machine learning (ML) models which use algorithms and techniques that accurately detect and stop muling as it’s happening. Regulatory pressures are encouraging more FIs to embrace real-time solutions. For example, since the Contingent Reimbursement Model rule change on October 7, 2024, UK-based FIs that receive scam funds now must split reimbursement costs 50/50 with sending FIs. These FIs are now incentivized to identify money mules and incoming APP fraud funds in real time to protect their customers and prevent illicit money from leaving their systems.

2: Leverage Supervised Machine Learning

Fraud prevention and AML efforts often rely upon unsupervised ML models to identify money mules. These models focus on identifying unusual or atypical patterns and perform inadequately given the complexity of money laundering and muling; after all, an unusual transaction doesn’t mean the customer is a money mule. These models fail to accurately identify money mules and lead to significant losses, while inundating analysts with false positives that contribute to alert fatigue and burnout. FIs need to use supervised ML models which train with labeled data given the complexity of digital transactions and crime. Supervised learning techniques outperform unsupervised learning and enhance model accuracy as the model is trained to identify money mules specifically, thus preventing illicit funds from flowing unabated and saving millions in mule losses. This also reduces false positives, alleviating analyst workloads and improving compliance operations.

3: Update Machine Learning Models Frequently

Mule detection models must quickly adapt and retrain to avoid drift and performance degradation given fast-changing criminal tactics, customer behaviors, and emerging products and technologies. Static models that retrain infrequently perform worse over time, detecting fewer mule accounts and generating more false positives. Daily adaptive models (DAMs), developed by Lynx, continuously update by retraining with new data and keep up with the latest trends in money muling. These models allow FIs to swiftly adapt to evolving criminal methodologies, payment technologies, and user behaviors.

4: Train Detection Models with Non-Transaction Data and Dynamic Payloads

Most mule detection models are trained exclusively with transaction payload data. While critical, transactions only tell part of the money muling story. Any customer account can become a mule account either knowingly or unknowingly at any point in time, making account, login, and onboarding data critical to understanding mule risk. In addition, most models only analyze transactions that match the payload type they were trained with and are unable to adapt to new types of data and payment channels. This is insufficient in the ever-changing payment system environment where new data fields and payment types constantly emerge and may be relevant to detecting money muling. Lynx’s DAMs integrate diverse data sources including customer account activities and transaction histories to refine their understanding of patterns associated with money mule activities. The models accurately distinguish between suspicious and legitimate transactions, stopping mules in their tracks and preventing illicit funds from leaving the FI’s systems without blocking genuine users. The models also use Lynx Flex, which enables dynamic payloads and incorporates new data types in model training. Lynx Flex additionally allows FIs to configure API and intelligence feeds through a no-code user interface, providing data extensibility that propagates to models, rules, and reports.

Read Lynx’s White Paper for More Insights

Interested to learn more about money mule techniques, global efforts to curb muling and money laundering, and how to implement cutting-edge detection solutions? Read our white paper Money Mules Revealed.