By Luke Housego Financial review
Payments giant Visa has applied artificial intelligence algorithms to the surging number of global e-commerce transactions it processes each year, and said the tech smarts have helped it reduce fraud-based losses to financial institutions by $US25 billion ($36.4 billion) in the last year.
Online payment fraud, or "card-not-present" fraud, occurs most commonly with the use of stolen card details and, according to Australian Payment Network's most recent report, accounted for 85 per cent of the total $561 million dollars worth of fraudulent transactions from Australian cards in 2017.
The problem impacts Australian banks as they often have to swallow the cost and refund customers for fraudulent purchases made with their accounts.
Visa's Asia-Pacific head of risk, Joe Cunningham, said the vast majority of fraud on payments cards worldwide was now taking place in e-commerce, and that applying AI to analyse more than 127 billion annual transactions meant suspicious transactions could be identified in real time.
"It is one of the earliest implementations of a neural network-based artificial intelligence [applications] in the world, apart from the military," Mr Cunningham said.
"If you take all the transactions that we help our clients identify as fraudulent, that comes to dollar figure of approximately $25 billion [USD] a year."
He said Visa's AI system has been evolving since the early 1990s and worked by scoring every transaction flowing across its network between zero and 99.
The score represents the probability of a payment being fraudulent and is shared in real time with Visa's clients.
"What's most beautiful about it is, the more it's used, the smarter it gets. So every single year, we pump hundreds of billions of transactions across the network, every single one of them is in the score," Mr Cunningham said.
New methods are also being used to reduce online payment fraud. "Tokenisation" replaces card numbers and other key information a unique digital identifier called a token.
Share the load
Rather than using "static" card details, the token uses dynamic information, such as a unique identifier of a known smart phone, allowing payments to be processed without exposing actual account details that could potentially be compromised.
Mr Cunningham said banks and financial institutions were also at the centre of fraud prevention, as an industry-wide approach.
"Ultimately it is up the bank to approve [a transaction] or decline it or choose to act in a different way," he said.
He said that while Visa has a significant data sample from its global payment network, the financial institutions had more detailed customer information, which increases the ability to identify fraudulent payments more accurately.
"In the ideal scenario, your bank is taking into account everything they know about you," Mr Cunningham said.
"Every single product that you have with them, whether it's a credit card, a home loan, a car loan, your deposits, your wealth products. They are hopefully taking everything into account when they look at your behaviour."