This is an excerpt from The Future of US Digital Payments 2025: ACH & Beyond.
Kinexys by J.P. Morgan’s Gloria Wan referenced the Trustpair
US Fraud Survey Report in conversation with Finextra, and mentioned that 90% of US companies experienced cyber fraud in 2024.
Nearly half (47%) of US companies also suffered a loss of over $10 million. Wan explained that there is “an opportunity for the industry to take immediate actions to enable stronger collaboration and greater data transparency to help businesses and individuals
manage fraud risks.”
As the volume, speed, and complexity of digital payments accelerates, as does the sophistication of financial fraud.
The financial services industry can no longer rely on responding to fraud risk; integration of proactive, real-time risk is needed to anticipate threats before they materialise. By preempting fraud, operational losses can be reduced and trust preserved across
the ecosystem.
Proactive, real-time risk detection is required at every stage of a transaction
Traditional fraud systems have historically been reactive and have only identified anomalies after they have breached security or resulted in financial losses.
Proactive risk recognition, on the other hand, allows payments firms to surface early signals of fraud, score and contextualise risk before a transaction completes, and model attack vectors with AI and ML to anticipate fraudulent behaviour across systems.
Carl Slabicki, executive platform owner, treasury services at BNY, indicated in conversation with Finextra, that fraud detection is where AI is making some of the biggest strides in payments. In addition to this, ML is currently used to spot suspicious patterns
before fraud happens.
“These models analyse massive amounts of transaction data in near real-time, flagging anomalies that humans might miss – and because they continuously learn, they adapt to new threats as they emerge,” Slabicki explored.
A single risk signal is not enough information to effectively and wholly mitigate fraud
In addition to taking a proactive approach, in 2025 and beyond, payments companies should prioritise leveraging information gathered from a range of risk signals and putting together a risk profile that fraudsters cannot bypass.
Techniques that should be used to mitigate fraud in the future include:
- Device intelligence: Emulators, rooted/jailbroken devices, multiple logins per device
- Behavioural biometrics: Typing cadence, tap velocity, navigation patterns
- Network risk indicators: Proxies, TOR nodes, IP reputation, velocity anomalies
- Transactional heuristics: Unusual transfer amounts, frequency spikes, payment rail anomalies
Recognising risk in real time allows for security measures that can be adapted according to perceived threat level. An intuitive user experience can be offered to low-risk users, while high-risk users are challenged – resolving the issue, yet not introducing
blanket friction across all transactions.
What does the industry think of the future of fraud prevention?
Renata Caine, general manager/senior vice president, embedded finance, Green Dot, agreed that a proactive fraud prevention strategy and investment in advanced technology allows payments players to have a competitive advantage. Further, she agreed that AI
and ML can improve anomaly detection.
However, she added that there is “no silver bullet for fraud prevention, but a multi-layer approach is the best option for mitigating fraud risk in digital payments.”
It has long been clear that collaboration is crucial to combatting fraud, and as Caine exemplified, there is “power in banks, fintechs and other industry players uniting to address and solve for this increasingly critical issue.”
She added that the “industry will never be able to stop fraud completely, as fraudsters are constantly evolving their tactics, but our chances of staying one step ahead of fraudsters to protect our customers are improved significantly through collaboration.”
After building up considerable intelligence on potential threats, payments companies should continuously retrain models, incorporate analyst feedback and measure false positives. A virtuous circle such as this allows each fraud attempt to strengthen the
system against future attacks.
On this point, Wan said that while “data sharing can deliver benefits for many use cases, the consideration on data privacy and legal constraints prevents sharing sensitive transaction and personal information data. With a federated machine learning mechanism,
the industry can collaborate beyond data sharing to enable insights and capability sharing while preserving data privacy.”
Moreover, Slabicki concluded by saying that as “fraud tactics become more advanced, the US digital payments sector will need to stay ahead by developing stronger ways to recognise risks before they cause any damage."