Kære læser. Artiklen her er en del af det engelske magasin Copenhagen Fintech. Indholdet er udformet på engelsk, da det også henvender sig til en udenlandsk læserskare, som deltager på eventen Money2020, hvor Berlingske Media er mediapartner. Magasinet er udformet af Berlingske Medias kommercielle redaktion i samarbejde med Copenhagen Fintech. God læselyst.
Sponsored by Nets
Never before has our payment card information been stored in so many different places as today. From the digital newspaper subscription to the online store, where you bought a pair of sneakers, card information is often stored in databases, with a risk of getting compromised.
In other words, criminals have an unprecedented amount of opportunity to steal card information and trade it on the black market among other things. For just two and a half years, monthly fraud attempts have tripled in Denmark from about 4,000 cards in 2014 to 12,000 cards in the summer of 2016.
"Criminals have become more aggressive and have been especially good at not just stealing card information, but also misusing them afterwards," says Kaspar Kock Kristensen, Senior Vice President for Fraud & Dispute Services at Nets.
Previously, criminal activity was mainly limited to misusing card information by manually buying expensive goods − such as TVs or software − in online stores and then reselling those products. Today, however, the techniques are much more sophisticated, according to Kristensen.
Using robot software, the criminals are now buying goods and services automatically in many different online stores at the same time. This way each store only sees a few transactions, which doesn’t look suspicious. Across several stores, however, the card may be used for 30 purchases a second.
“The technologies behind card fraud have become more advanced and that compels us to get even better at preventing the fraud from happening,” Kristensen says.
For the last couple of years, Nets has been working hard to do just that by rolling out its latest weapon against card fraud: artificial intelligence. The efforts are paying off, it seems.
Automating fraud prevention
Detecting card fraud before it happens can be rather complex. There are, of course, the most obvious cases, when a card is being used at a physical store in Brazil and then five minutes later in China – something a regular shopper would never be able to do. Analysts at Nets make logical rules to block payments that fit these kinds of usage patterns. Yet, the challenge grows when criminals find new and clever ways to implement their fraud schemes using automated software. Nets have already developed over 700 different logical rules that block fraudulent behaviour – but they are getting increasingly complex to maintain and finetune.
“We have to adapt our systems to detect new patterns of fraud at an ever faster pace. This all points towards using artificial intelligence in combination with huge amounts of data if we are to catch up with them,” Kristensen says.
As one of the leading anti-fraud services providers in the Nordics, Nets is handling payments from over 19 million cards. This gives them a unique insight into the behaviour and patterns that explains not only how criminals abuse the system but also the way each of us buy things.
Payment data is now being used in a neural network, which is a way of artificial intelligence, to detect patterns in a complex system. It understands the unique payment behaviour of each cardholder and triggers an alarm when something falls outside of normal spending patterns.
»Our goal is to make sure that legal users are unaffected.«
Combined with the logical rules, the neural network allows Nets to detect a larger number of fraudulent payments even before they happen.
“The system works in real time and allows us to immediately reject a payment that looks suspicious,” Kristensen says.
Since the implementation of fraud prevention with neural network capabilities, fraud attempts have already started to lessen. Monthly fraud attempts went down again from 12,000 cards last summer to 8,000 cards in the first quarter of 2017, according to Nets.
“We prevented an additional 150 million kroner worth of card fraud in 2016 compared to 2014 in Denmark alone,” Kristensen says.
The white noise challenge
Detecting and stopping criminals is not the only difficult task fraud prevention faces. It is also supposed to avoid blocking payments that look suspicious but really are genuine cardholder payments − a concept known as white noise.
Perhaps you normally use your debit card to pay for groceries and other small purchases, but what about the moment you decide to buy an expensive holiday online for the first time? The neural network might consider it an anomaly, block the payment and ultimately block the card.
Because of scenarios like this, Nets still uses human agents to decide if a payment card should be cancelled.
“Our goal is to make sure that legal users are unaffected,” Kristensen says while explaining that achieving this gets easier with more data: “Our agent performance is already three times as good as when we started, because we now have a more detailed insight into how they perform and which rules work and which don’t.”
In the end, businesses, banks and consumers all end up saving money, when fraud prevention improves on a large scale. When a card gets misused, the merchant covers the bill in 80-90 % of the cases, while the bank covers the rest. Additionally, banks and me rchants have administration and dispute handling expenses for each fraud case.
This highlights the issue of online payments, where around 80 percent of fraud cases happen, according to Nets. On the Web, security can be scarce and payments are often authorised by their users simply typing in the card number.
This is gradually changing as more online stores implement the 3D Secure standard, requiring consumers to type in a pin sent to their mobiles before paying.
“Regular consumers shouldn’t be too worried about fraud, since it only happens in about 0.05 percent of the cases,” Kristensen says. “However, I still believe, that criminals will continuously force us to keep improving our technology to be able to prevent fraud in the years to come.”
This article is part of the commercial publication 'Copenhagen Fintech'. Click here to view all articles