Using AI to tackle fraud

Elavon has a strong track record in delivering flexible and scalable security solutions for payment systems. Building on that track record, we spent the past two years developing the next generation of fraud-detection capabilities powered by artificial intelligence (AI). 

The resulting advanced solution, Elavon Pay Defence combines machine learning and link analysis and launched in 2025.

Candice Pressinger, Director of Customer Data Security at Elavon Europe, said: “AI isn’t a silver bullet. It’s a tool that needs human oversight, good data and constant evolution.  

“We’re using AI to identify anomalies in transaction behaviour in real time, helping retailers to catch fraud before it happens. 

“The trick is to use AI as a tool, not a crutch. To stay ahead of the fraudsters, it needs to be paired with robust data, vigilant oversight, and clear human decision-making.”

Globally, machine learning tools are increasingly used for fraud detection. In 2023, 48% of merchants were using them, and 66% of those surveyed found them effective in the fight against fraud, according to SlashNext’s ‘The State of Phishing’ report.

“While this reflects a proactive response to the growing threat of online fraud, it also highlights the magnitude of the challenge faced by merchants in safeguarding their businesses,” says Candice. 

Machine learning models to analyse patterns

Informed by huge amounts of transactional data, machine-learning models can identify intricate patterns and anomalies that indicate fraudulent behaviour with incredible accuracy. 

Beginning with historical data and incorporating company policies and risk tolerances, data points that are important to decision-making combine into machine-learning features. 

All of this culminates into a model which will accept, review or block recommendations for each customer or each transaction. 

“Our latest machine-learning approach to fraud prevention is not rules-based. It analyses patterns in real time and continuously learns, grows and changes. Two merchants in the same retail category won’t be the same in terms of the ‘back doors’ fraudsters’ use to break into their systems – and fraudsters are using AI to create ever-more-convincing scams to identify these back doors,” Candice explains.

“For many businesses, the ideal setup combines AI with human oversight, with the latter occasionally required for manual review, as well as setting fraud rules where recommended,” Candice explains. 

“By continuously refining their models based on real-time feedback, these algorithms empower merchants to stay ahead of evolving fraud tactics and mitigate potential losses before they happen.” 

Link analysis providing a holistic view

Working alongside the machine learning engine within Elavon Pay Defence, link analysis is a technique that examines the intricate web of relationships between dozens of data points across the entire clientele to offer a holistic perspective on potentially fraudulent activities.

“Unlike legacy fraud detection methods, which tend to focus on individual transactions, link analysis enables merchants to uncover hidden connections and identify fraud networks operating across multiple accounts and transactions,” says Candice.  

“This means we’re no longer just looking at cards or cardholders in isolation and are looking at all shoppers in a holistic way. In a time when ecommerce fraud continues to grow in scale and sophistication, unravelling these complex networks, merchants can pre-emptively thwart fraudulent activities and safeguard their businesses against financial losses and reputational damage.

“In addition, link analysis can be used to inform both machine-learning features and human-created rule sets, providing further automation and convenience.” 

While the ecommerce fraud epidemic poses a threat to businesses and consumers alike, by harnessing the power of these advanced technologies, merchants can detect and prevent fraudulent activity of various types.

“Merchants can reduce costly chargebacks, improve first-time authorisation success, and ultimately boost revenues with greater accuracy than ever before. Ultimately, it means they can boost their revenues with unparalleled precision in the dynamic landscape of online commerce,” says Candice.