Fraudsters continue to pose huge issues for the financial sector across the globe. UK Finance recently revealed that criminals stole over £1billion in 2023 alone. Even worse, the fraud landscape shows no signs of improving, as bad actors increasingly utilise AI to increase the damage done to financial organisations. However, many of the largest financial organisations worldwide are actively investigating measures to counteract these threats.

Financial crime continues to grow, seemingly exponentially, with fraud damages expected to hit $10.5trillion annually by 2025, a drastic rise from $3trillion in 2015.

Meanwhile, the fraud landscape is rapidly changing. Over a third of fraud attempts (42.5 per cent) targeting financial institutions now use AI, according to a recent study by digital identity and fraud prevention solution Signicat. Overall, around 29 per cent of these AI-driven fraud efforts are successful.

Even if these success rates remain at around the same level, the sheer volume of attempts mean that fraud levels could ‘explode’ the firm has warned. As part of its study, Signicat also found that many organisations are largely unprepared to evolve their approach to counter the uptake of this threat.

However, not all financial institutions are sitting still. Many of the world’s largest financial bodies are actively investigating ways to stamp out AI-driven fraud with one specific emerging technology AI.

To find out how efforts of pitting AI against AI are taking shape, and how this could evolve in the future, we take a look at some of the latest anti-fraud approaches utilising AI.

Pay.UK’s anti-fraud pilot

Pay.UK, the operator and standards body for the UK’s retail interbank payment systems, has now revealed the results of its AI-driven fraud detection and prevention pilot, in collaboration with Visa, Synectics Solutions and Featurespace.

The standards body confirmed the pilot in June 2023, after contacting industry partners to test the benefits of the service with a group of participating banks and payment service providers. It ran for three months and trailled a new overlay service, enabling all UK banks and building societies to analuse money flows and use predictive intelligence to detect fraud and prevent crime before it occurs.

Kate Frankish, chief business development officer and anti-fraud lead at Pay.UKKate Frankish, chief business development officer and anti-fraud lead at Pay.UK
Kate Frankish, chief business development officer and anti-fraud lead at Pay.UK

Following extensive testing, the pilot produced an average 40 per cent uplift in fraud detection, with a 5:1 false positive rate. This would equate to over £112million worth of fraud detected annually.

Kate Frankish, chief business development officer and anti-fraud lead at Pay.UK, discussed the success: “The positive results from this pilot demonstrate the importance of innovation and cross-industry collaboration in developing effective solutions to stay ahead of fraudsters and protect people in the ever-changing payments landscape.

“In 2023, the UK saw 232,429 people falling victim to fraud. To reduce the scale of the crime that is happening we need a unified approach, and this future service will be a major step forward.”

Visa takes matters into its own hands

As part of this pilot with Pay.UK, payments giant Visa analysed billions of UK account-to-account transactions, correctly identifying an additional 54 per cent of fraud and APP scams beyond those identified by the banks’ own fraud prevention systems.

It did so by leveraging the latest AI technology, proving that using predictive AI technology could potentially save £330million for UK consumers, businesses and the economy.

Now, Visa is making this real-time fraud detection service, dubbed ‘Visa Protect for A2A Payments’, available to all banks in the UK. This new technology aims to help intercept suspected fraudulent transactions in real-time, stopping scams before any money ever leaves a victim’s bank account.

Mandy LambMandy Lamb
Mandy Lamb, managing director at Visa UK & Ireland

Mandy Lamb, managing director at Visa UK & Ireland, commented: “The UK has one of the most developed payment systems in the world, but also sees some of the highest levels of account-to-account fraud. Once fraud happens, the money is in the hands of the criminals so fraud prevention must be our collective goal, in the financial services industry and beyond.

“Visa has already reduced card payment fraud to historic lows, which we are very proud of. We are now bringing our AI capabilities to fight fraud and scams on account-to-account payments before they happen. We are really excited about working with our partners on this – keeping people and businesses safe from scammers is the biggest priority for Visa.”

Unleashing AI’s potential with collaboration

Testing also continues to progress across other world-leading organisations. Swift, the global financial messaging service, is set to launch two pilots of its own which will test the practical application of AI to enhance fraud detection in payments.

Swift is piloting a new enhancement to its existing Payment Controls Service, which enables financial institutions to flag or block anomalous payments before they are made. The pilot will involve five Payment Controls customers, including India-based Axis Bank, to test a new approach that uses AI-based algorithms to help them better detect fraud in transactions.

It explained that it will train the new AI model using historical patterns of activity on the Swift network to create a ‘more nuanced and accurate’ picture of potentially fraudulent activity.

Tom Zschach, chief innovation officer at SwiftTom Zschach, chief innovation officer at Swift
Tom Zschach, chief innovation officer at Swift

Swift’s second pilot is focused on collaboration and will leverage its established position to help financial institutions share insights to improve fraud detection worldwide. This pilot will run with involvement from the likes of BNY Mellon, Deutsche Bank, DNB, HSBC, Intesa Sanpaolo and Standard Bank.

“AI has great potential to significantly reduce fraud in the financial industry. That’s an incredibly exciting prospect, but one that will require strong collaboration. Swift has a unique ability to bring financial organisations together to harness the benefits of AI to help further strengthen the cross-border payments ecosystem,” explains Tom Zschach, Swift’s chief innovation officer.

How does AI-driven fraud prevention work?

Everywhere you look, AI-related innovations are taking place. To understand how firms are implementing AI, and the complexities involved in that, we spoke to Ariel Shoham, vice president of risk product at Mangopay, a modular and flexible payment infrastructure provider for platforms.

Mangopay launched its own payment processor-agnostic AI-driven fraud prevention solution earlier this month, which hopes to tackle the account takeovers reseller fraud, payment fraud, chargebacks, and return abuse.

Ariel Shoham, vice president of risk product at MangopayAriel Shoham, vice president of risk product at Mangopay
Ariel Shoham, vice president of risk product at Mangopay

“AI in fraud prevention is the key element that complements all the other fraud detection actions by increasing precision and automating the system for real-time results,” Shoham explained.

He also broke down how Mangopay leverages AI: “Our risk detection system gathers thousands of data attributes about the users’ devices, networks, and behaviour, and identifies numerous platform-specific risk signals, also leveraging dark web insights.

“The next step is to process all these data with AI. Our ML models spot non-obvious patterns and automatically eliminate potential fraud threats which present known MO signals, or deviations from expected normal patterns. Machine learning takes historical data to help us establish the alerts, and in borderline cases, our data science team takes a closer look for further investigation.

“The last step is the decision-making process. The key differentiator here is explainable AI – a transparent decision engine that shows why a transaction is accepted or refused through clear explanations of our risk signals. This clarity is essential for platforms to understand the rationale behind flagged activities and refine their anti-fraud strategies over time and avoid unexplained ‘black box’ biases.

What are the biggest challenges when working with AI?

Finally, Shoham explained the most difficult aspects of fighting fraud using AI: “Consolidating our Fraud Prevention product involved some complexities which are typical of those faced by fraud prevention providers. Fraudsters evolve their tactics to bypass risk detection, so our team must stay tuned in real-time to ensure the product stays at the top of its game.

“Secondly, with the roughly two million transactions that we monitor on a daily basis and billions of transactions to process overall, analysing this massive amount of data can be challenging. In terms of infrastructure, we needed to develop a system capable of scaling and managing the increasing load efficiently, both during setup and ongoing operation.”

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *