How to Find “Money Mules” with Machine Learning Fraud Detection Software

As more financial institutions are cracking down on fraud, criminals are turning to more creative measures to conduct illicit transactions and other activities. One common method on the rise is using so-called “money mules” to help facilitate transactions in hopes of keeping their money movements under the radar. 

COVID-19 has further increased the demand for money mules as crime never takes a holiday, even during a crisis.

In response, financial institutions should explore the advantages of machine learning fraud detection software that can help identify potential money mule activity.

What are Money Mules?

The term “money mule” refers to a person who is used to transfer money in an effort to conceal financial activity. For example, someone who wants to illegally transfer ill-gotten money to someone else may use a money mule to complete the transaction and make it look legitimate. The goal is to add extra layers to the money trail to make transactions look less suspicious.

To date, money mules have been a relatively safe bet for criminal activity, as many money mules don’t realize they’re being used to launder money. The originator will wire money to the money mule, then the mule will convert that money into cash or a check, wire it into a third party bank account, or convert it into a prepaid debit card, among other options.

On the surface, this type of activity seems harmless. Millions of people request cash, write checks, or buy prepaid debit cards on a daily basis, so it’s not always easy to detect the work of a money mule. However, given that the number of money mule cases are on the rise (40,000 cases reported in the UK in 2018 alone), it’s no surprise that financial institutions are ramping up efforts to detect and eliminate this type of fraud. 

The Hallmarks of a Money Mule

The money mule profile takes several forms, primarily young or middle-aged men, social media users, and students. 

Money launderers prefer targets that have no criminal history to lower their risk of getting caught. They hope that the lack of criminal activity will draw less attention to the transactions. 

Another sign is unusual bank activity, such as sudden spikes and withdrawals in quick succession. Though one large deposit and rapid withdrawal may not indicate money mule activity, several of these transactions could be more revealing, providing a financial institution’s tools are designed to look for these patterns.

Scammers are also getting more creative with their approach to recruit money mules. Recently, one report noted that a crime ring disguised as a COVID-19 relief effort was hiring people to collect and transport funds. The website for the group imitated another non-profit and had registered the website just weeks prior, even though they claimed to have been in business for years. 

In addition, money mule schemers are increasingly turning to messaging apps and social media channels to attract interested parties, often in the form of “get rich quick” schemes. The money mule will receive payment in their bank account, then be instructed to send the funds to another account or take out the money and give it to someone, all while keeping a cut of the funds for themselves.

It’s also important to note that money mule activity isn’t always conducted in large transactions. For example, several $10 or $100 transactions can easily rack up millions of dollars in money laundering when spread out over hundreds of money mules and accounts. Small amounts and subtle behaviors help to keep them under the radar, and many banks are none the wiser. 

For many, the offer’s lucrativeness is an easy sell. There’s little work involved and the pay could be substantial. College students are common victims, as they need a way to support themselves while in school. 

Many do not realize they’re doing anything illegal, but if caught, they could face up to 14 years in prison.  

How to Spot (and Stop) Money Mule Activity with AI and Machine Learning

Because money mule activity looks normal at face value, many financial institutions are at a loss as to how they can identify and prevent these illicit activities. 

AI and machine learning are among the few tools designed to tackle the money mule challenges that go beyond reviewing individual transactions. 

In a typical review process, transactions are viewed individually and usually look legitimate on the surface level. However, today’s attacks are carefully coordinated and require a big picture understanding to detect patterns and illicit behaviors. 

AI and machine learning can review transactions on a much larger scale and pick up on trends that would elude the average manual review process. For example, a person testing a transfer with a small amount and then canceling it before sending $600 to another individual could hardly be considered suspicious. However, fifty people testing a small transfer and canceling it, then sending $600 to the same individual around the same time might raise an alert.

DataVisor’s platform works by looking beyond the single customer and their transactions and reviewing the connections between customers and transactions to pick up on trends and patterns. AI and machine learning are able to do this at scale, where it would take teams countless hours and manual reviews to map just a fraction of what DataVisor could do in real-time. 

Many banks are already using rules setup and supervised machine learning to detect fraud patterns, and these are good methods for picking up on known fraud. However, fraud conducted by money mules is difficult to detect and may not follow typical patterns. Data collected by supervised machine learning requires labeling, which can be time- and resource-intensive, and this prevents new attacks or patterns from being detected until data is correctly stored and labeled.

The best approach is to leverage unsupervised machine learning (UML), which doesn’t require training or labeling and provides early detection of potential fraud. UML isn’t restricted to preset rules and can link various data sets to deliver deeper insights. 

Combining UML with supervised machine learning and rules gives financial institutions a complete solution to detect known and unknown fraud. Early identification is critical to solving money mule challenges and stopping fraud in its tracks.