top of page
  • Writer's pictureRoy Urrico

Q2 Issues Guidance on Thwarting Check Fraud Using Machine Learning, Artificial Intelligence

Source: Q2's "Put a Hard Stop on mRDC Check Fraud" guide

By Roy Urrico

Finopotamus aims to highlight white papers, surveys, analyses and reports that provide a glimpse as to what is taking place and/or impacting credit unions and other organizations in the financial services industry.

Fraudsters have become more proficient at creating near-perfect counterfeit checks, entering them via mobile remote deposit capture (mRDC) into transaction and business processes, and transferring funds out of consumer and businesses accounts. That is a message delivered by Put a Hard Stop on mRDC Check Fraud, a guide from Q2 that provides ideas on using machine learning to mitigate check fraud in mobile banking.

“Left with negative balances and lots of stress, mRDC fraud is a bad enough experience to make accountholders consider leaving their FI (financial institution),” reported Austin, Texas-based Q2, which provides digital banking and lending solutions to banks, credit unions, alternative finance companies, and fintechs in the U.S. “One way to ramp up vigilance is through machine learning. This guide will help your financial institution in assessing the benefits of machine learning in further mitigating check fraud in mobile banking.”

The Current State of Check Fraud

The report maintained checks continue to be the payment method most often targeted by fraudsters. “In 2022, the Association of Financial Professionals (AFP) found that two-thirds of organizations were prey to check fraud. Checks are the payment method most used by businesses, and so not surprisingly are the most frequent targets of fraudsters.”

The report pointed to how, in 2023, “check fraud continues to be a critical concern to financial institutions, with many describing fraudsters as rampant and growing even more brazen. For an FI hit by fraudsters, not only can a bank or credit union lose customers and members, but the reputational damage can hinder their growth.”

The report also noted tampered and forged checks are now sold on the dark web. “That poses an even more concerning picture and points to the need for more innovative technology to stop the higher volume that can come from these underground markets.”

The Q2 guide also noted that concurrent to these fraudulent activities taking place is the rise of mobile-first banking as the preferred choice for accountholders. “Various statistics point to over 50% of consumer accountholders using mobile banking to transfer funds or pay bills, and the pandemic has created a shift to more businesses using mobile-first banking,” said the report.

As a result, this greater increase in mobile and mRDC use increases the volume of checks that FIs need to assess for fraud. “Costly slip ups are practically inevitable — especially if anti-fraud technology isn’t in place to help.”

Not Able to Obstruct Check Fraud

Q2 referred to AFP research that said: “While treasury and finance leaders are very aware of how widespread this type of fraud has become, they are not able to obstruct it sufficiently.”

Fraudsters are successfully infiltrating payment activity at organizations by using email to do so, AFP noted. “Their success in deceiving organizations encourages them to continue to use BEC (business email compromise).” In BEC the invader obtains access to a corporate email account and spoofs the organization’s identity to scam the company or its employees, clients or associates.

“Machine learning (ML), or artificial intelligence (AI), can lead to much better control of check fraud by giving a bank or credit union another weapon in its fight,” the guide suggested. “Taking previous algorithmic models used in detecting other banking activity anomalies in real time, Q2 is using them to significantly reduce check fraud in mobile banking.”

Q2 Chief Data Scientist Jesse Barbour .

As Q2 Chief Data Scientist Jesse Barbour described, “We are training models that solve problems in the computer vision domain. Toward this end, we work closely with our customers to identify the complex and subtle features in check images that can be useful in detecting fraud.”

Barbour added, “Leveraging techniques from the ML computer vision domain, what can be difficult for the eye to pick up, can be detected with greater accuracy and at a scale unattainable by humans.” He added, the approach in development at Q2 requires deep learning and it can be a challenge mathematically, not to mention connecting a neural network together to bring a cohesive knowledge into place.

The guide described how Barbour and fellow Q2 Data Scientist Kristi Voll are working with FIs that want to make an impact in preventing check fraud. Their collaboration is making significant headway in studying check visuals and determining images to train a neural network.

Voll pointed out, “Existing ID systems are not catching all these fraudulent checks and our work is focused on blind spots to those engaged with mRDC. Computer Al will recognize these and this all happens as checks move through systems.”


bottom of page