Today, there are few forms of financial fraud that are more prevalent – or more costly – than credit card fraud. It’s difficult to establish an exact accounting of the losses incurred each year due to credit card fraud, but one report put the cost at $21.84 billion in 2015. Troublingly, it’s a problem that seems to be getting worse, not better, at least in the United States.
According to the Federal Trade Commission, reports of credit card fraud went up by 23% last year alone, with no signs of slowing down. It’s increasing despite a steady advancement towards more secure cards and transaction methods, leading many of the world’s largest merchants and credit card issuers to search for new solutions to the problem. That’s why more industry players are turning to AI and machine learning techniques to limit or prevent fraudulent activity. Here’s a look at where the technology stands now and how AI is joining in the fight to end credit card fraud once and for all.
A fragmented system
Part of the difficulty with preventing credit card fraud is the fact that it’s a global system with no unified standards. For example, in the U.S., merchants and processors notoriously feuded for years about who was going to foot the bill for the rollout of EMV-enabled cards and processing equipment. Then, when government regulations forced their hand, the industry deployed a version of the system that watered down the inherent security benefits of EMV cards.
While the industry dithered, another trend took hold: e-commerce. Each year, an ever-larger slice of global retail sales move online, where EMV cards offer no protection. While some purchasers opt for the security of a virtual credit card, most do not. That shift has been a boon to fraudsters, who have targeted online retailers to steal consumer credit card data en masse. It has also forced retailers and credit card processors to explore AI as a means of finally solving the industry’s costliest problem.
A helpful dialogue
One of the best examples of how AI is beginning to see use as an anti-fraud measure in the credit card industry is Eno, the AI-powered virtual assistant offered by consumer credit giant Capital One. The system uses a combination of natural language processing and pattern recognition technology to facilitate real-time communication with customers. When a potentially fraudulent transaction is flagged by the system, the user receives an instantaneous query from Eno asking if they’re aware of it. Eno then analyzes the user’s responses, both to further personalize the exchange and to look for patterns of deception. That makes Eno effective at spotting friendly fraud too, which has been on the rise of late.
Looking for the exception
Retailers have also been pushing the development of AI fraud detection techniques. For them, the problem is arguably harder to approach than it is for credit card issuers since they lack access to each customer’s whole purchase history. To overcome that difficulty, eBay has turned the problem on its head by looking less at identifying patterns that indicate fraud and more at patterns of normal customer activity. In a paper released recently, eBay’s researchers propose a so-called “outlier detection” AI system. They theorize that it’s easier for a retailer to train an AI to know what a legitimate transaction looks like on their platform (since they have access to that data) and then search for signs of abnormal behavior. In testing, the system can already spot fraudulent purchases with high precision 40% of the time with no direct human intervention. As a first effort, that’s a spectacular success rate – and one that should improve as the system gains real-world experience.
Securing all sides
Given the rapid advances being made in AI-powered fraud detection systems by merchants and credit card issuers, it’s easy to foresee a future when would-be criminals have little chance of slipping through the cracks in the system, despite its fragmented nature. As the latest AI solutions start to spread through all of the stakeholders in the consumer credit ecosystem, the result should be a multi-layered anti-fraud system with multiple checks on every transaction in real time. It’s also important to note that all of this will be completely behind the scenes, so there will be no noticeable impact on consumers. If these developments continue apace, there’s a good chance that the inter-industry arguments over fraud liability will soon be a thing of the past – and that’s good news for everyone.
To learn more about how AI is spreading throughout every industry, read “The AI Gold Rush: Artificial Intelligence And Machine Learning.”