Banking Exchange Magazine Logo

AI and Machine Learning as a Solution to the EBT Fraud Epidemic

Fraudsters can leverage several different methods to steal SNAP/EBT benefits

  • |
  • Written by  Ali Solehdin, Chief Product and Strategy Officer at INETCO
AI and Machine Learning as a Solution to the EBT Fraud Epidemic

In 2022 alone, the Supplemental Nutrition Assistance Program (SNAP) distributed over $113.9 billion to nearly 22 million households across the United States. This figure represents an increase of over $5 billion from the year prior and nearly a $40 billion increase from 2020. Unfortunately, as the SNAP allocation has increased, criminals — from individuals to organized crime rings — have stolen an increasing share of these benefits. In fact, EBT fraud is now estimated to be costing taxpayers up to $4.7 billion annually, according to the Government Accountability Office.

Since its inception in 1939 the Food Stamp Program, known today as the SNAP, has been a vital resource for Americans near or below the poverty line. The program was significantly modernized in 1984 when the first Electronic Benefits Transfer (EBT) pilot was launched in Reading, Pennsylvania to digitize these types of benefits payments. Initially, EBT helped curb food stamp fraud by creating an electronic record of each food stamp transaction, which made it simpler to identify violations. However, as digital payments technology has continued to evolve since that time, so too have the methods employed by fraudsters and scammers working to separate the most vulnerable segments of our population from their much-needed benefits. This has led to a criminal epidemic, with billions of dollars of SNAP/EBT benefits payments now being siphoned away from their intended recipients through card skimming and other forms of fraud on an annual basis.

Fraudsters can leverage several different methods to steal SNAP/EBT benefits. Card skimming has become rampant and is achieved by installing a skimming device on an ATM or POS system to clone cards and PIN numbers. EBT cards are highly antiquated and only contain a magnetic stripe rather than an embedded EVM chip, which makes them particularly vulnerable. While bulk purchases of magnetic stripe cards lowers costs in the short term, they are far more susceptible to forgery and make a much easier target for fraudsters. In addition to card skimming, criminals can also steal EBT funds through account takeovers, with a wealth of data and ill-gotten account information available through channels like the Dark Web.

One solution to the uptick in EBT fraud we’re seeing would be to simply overhaul the SNAP/EBT program to ensure chips are included in all benefits cards. While this is currently under consideration by several states, it would take several years to implement. Even if this was achieved in a best-case scenario time frame of around two years, states could still be susceptible to $9 billion more damage in EBT fraud during that window. While that is a grim thought to be sure, what’s perhaps even worse is that by the time this overhaul could be completed, it’s more than likely that criminals will have found new methods through which to extract SNAP funds. So, what is the solution? States and their SNAP/EBT distributor partners need to rapidly deploy fraud prevention technology that can detect anomalies in real-time, bolstered by AI and machine learning that can learn on the fly and adapt even quicker than criminals can. Thankfully, this technology already exists.

AI-driven machine learning models can create individual profiles for each EBT card that is created. Then, these systems can apply behavioral analytics to automatically detect fraudulent activity related to these cards in real-time. This technology can also train itself to evolve over time to detect new schemes and forms of payments fraud and does not require data scientists or deep technical expertise to operate. This is a far more efficient means of detecting, deterring and mitigating SNAP/EBT fraud than replacing millions of benefits cards or setting up sting operations to capture criminals, both of which are extremely costly and require enormous manpower. Real-time fraud prevention tools underpinned by AI and machine learning can also be deployed in a mere 60-90 days and rapidly deliver a significant reduction in the volume of EBT fraud taking place. Applying this type of solution would have the potential to save taxpayers billions of dollars a year, while also freeing up significant funds and resources for states and government agencies to direct to other areas of need.

Protecting our communities’ most vulnerable citizens is paramount. The epidemic of EBT fraud has gotten completely out of hand but it’s not too late to significantly curb this criminal activity. Technology has given us the tools to track far more data points associated with each EBT card to prevent skimming and card fraud in real-time with more effectiveness than we could ever hope to do manually. While AI and machine learning have almost become buzzwords as of late, these tools have boundless potential to make our world a safer, more hospitable place and leveraging them to exponentially reduce EBT fraud can be a true gamechanger for both individuals and states across America.

Ali Solehdin is the Chief Product and Strategy Officer at INETCO; the company’s BullzAI platform is a preferred tool used by government agencies and payment administrators to detect and block fraudulent EBT transactions due to card skimming and account takeovers.

back to top


About Us

Connect With Us



How to get the most out of Data and AI
with Ravi Loganathan from Sardine
and President of Sonar

Wednesday, July 24, 2024 at 11 AM ET / 8 AM PT

In this webinar we will cover:


This webinar is brought to you by:

SardineBanking Exchange