Dan Dica, CEO of Lynx, discusses how Artificial Intelligence could be the key to preventing cases of fast fraud.
The payment landscape has indisputably altered in the financial sphere, where instantaneous transactions have become the norm. Cash is no longer king, and the speed with which money moves across borders is unprecedented. Investment in the payment space from major tech entities and venture capitalists is a testament to the fact that payments support and propel digital economies into the future.
However, with swift transactions comes the spectre of accelerated fast fraud, challenging financial institutions to strike a delicate balance between speed and security. We find ourselves at a juncture where quick payment mechanisms are essential, yet they must be fortified with robust safeguards to protect against ever-evolving threats.
The changing face of fast fraud
As payments have evolved, so too have the tactics of fraudsters. With our behaviours changing rapidly – as demonstrated by the unstoppable rise of e-commerce – the battleground is shifting, too. The vast majority of fraudulent activity now starts outside the banking sector, with over three in four (77%) of fraud cases in the UK originating from online sources.
Today’s financial criminals are savvy, leveraging low-cost tools that exploit recent technological advancements – such as Generative AI. Synthetic identity fraud attacks are increasing rapidly, while social engineering scams – such as Authorised Push Payment Fraud (APPF) – have grown significantly more sophisticated.
APPF happens when the attacker socially engineers a victim into transferring money to an account the attacker controls. Many of us have received suspicious texts over the past year in which fraudsters act as banks, contractors, or even family members to convince us to move money from one account to another – unfortunately, many of these attempts are successful. In the first half of 2023 alone, £239.3 million was lost to APPF in the UK, with just 64% of the total loss returned to the victims despite the good intentions of financial institutions.
To add to these challenges, automated synthetic identity fraud is becoming a big challenge, resulting in mass mule accounts following the compromise of identities. These accounts are used by criminals to wash money. According to Experian, 42% of first-party current account fraud is now mule-related, with the fraud rate for current accounts rising by 13% in the first three months in the UK in 2023.
This new approach taken by attackers can leave a financial institution vulnerable to a network of mule accounts, sometimes actively used in money laundering, other times staying dormant and waiting to act at a moment’s notice.
The urgency of fast payments leaves little time for due diligence. As speed increases, there is a shrinking window for detection and interdiction before fraudsters receive (and often immediately withdraw) large sums. It’s clear that change is needed.
Smarter prevention over knee-jerk reactions
Fraud prevention has relied heavily on analysing past threats and rule-based filters. This has left institutions often reactionary and exposed, struggling to keep pace with an evolving landscape. The rules in question do not automatically learn, are unable to predict new types of attacks and – crucially – the attack must be the same as the rule that’s defending against the attack, meaning that rule writers often have to script tens of thousands of scenarios. This is an archaic practice. In 2024, effective prevention demands predictive, adaptive defences.
Artificial Intelligence (AI) is emerging as a key asset in the fightback against fraud. Coupled with bespoke algorithms, AI can gain an understanding of an individual’s behaviours and spending – tracking everything from transaction behaviour to location. Rather than inflexible and often outdated rules that block transactions, these solutions instead facilitate precise decisions that minimise false positives while maximising security. The best-in-class solutions are self-learning Machine Learning models that are trained daily. The daily training should include data from the day and the relevant characteristics and instances of confirmed fraud. This ensures that the machine learning models adapt to changes in customer behaviour attacks and new types of fraud.
In today’s world, where speed and convenience are deeply valued, the more that can be removed from human decision-making – especially in potentially stressful situations – and instead automated under a predefined risk matrix, the better. As payments accelerate, so too must fraud prevention to ensure transactions happen safely. AI and machine learning are critical weapons to get ahead of threats in real time without introducing friction. We can protect customers by leading smart technology while ushering in a new era of payment innovation worldwide.
