Fraud Risk Mitigation in Real-Time Payments: A Strategic Agent-Based Analysis

Katherine Mayo, Nicholas Grabill, Michael P. WellmanIJCAI 2024 · OpenReview

Abstract

Whereas standard financial mechanisms for payment may take days to finalize, real-time payments (RTPs) provide immediate processing and final receipt of funds. The speed of settlement benefits customers, but raises vulnerability to fraud. We seek to understand how bank nodes may strategically mitigate fraud risk in RTPs, through investment in fraud detection and restricting payments eligible for real-time processing. To study this, we introduce an agent-based model of the payment network supporting both real-time and standard payments, and define a game among banks and fraudsters. Using empirical game-theoretic analysis, we identify Nash equilibria in nine game configurations defined by network attributes. Our analysis finds that as banks become more liable for fraud, they continue to allow RTPs but are more likely to employ both restrictions and a high level of fraud detection. Fraudsters, in response, switch from targeting only RTPs to attempting fraud with any type of payment and tend to exploit banks where they have historically been most successful. We also conduct a strategic feature gains assessment to further understand the benefit offered by each of the bank's risk mitigation measures, which confirms the importance of selective RTP restrictions. Finally, we find that in equilibrium bank strategic decisions negatively affect fraudsters while minimally impacting customers.

Introduction

Traditional bank transfers often involve a multi-day "buffer" that allows for manual fraud detection and reversal. The advent of Real-Time Payments (RTPs), such as Zelle or FedNow, has eliminated this buffer, settling transactions in seconds. While this provides significant utility to consumers, it creates a "speed vs. security" paradox for financial institutions. Without the ability to manually review suspicious transactions, banks must rely on automated, strategic defenses to combat increasingly adaptive fraudsters.

This paper explores the optimal balance of risk mitigation strategies in a network where both banks and fraudsters act strategically. By employing agent-based modeling and game theory, we analyze how banks can protect themselves and their customers without compromising the efficiency of modern payment systems.

Modeling the Payment Ecosystem

To analyze these complex interactions, we developed a simulation environment utilizing two primary frameworks: Agent-Based Modeling (ABM) and Empirical Game-Theoretic Analysis (EGTA).

Financial Credit Networks

The simulation tracks the movement of funds using Financial Credit Networks (FCNs). In an FCN, edges represent credit and debt obligations between nodes. This allows the model to accurately simulate how money moves—or is delayed—between customers and banks across an interbank network.

A basic Financial Credit Network (FCN) representing the credit and debt relationship between a customer and a bank.

Agent-Based and Adversarial Dynamics

The model consists of 200 customer nodes, 4 bank nodes, and a strategic fraudster node. Customers possess balances drawn from an exponential distribution (averaging 10,000 units) and engage in transactions up to 1,000 units. The fraudster node is designed to be adaptive, utilizing a Multi-Armed Bandit approach (specifically the Upper Confidence Bound algorithm) to target banks where it has achieved the highest historical success.

The agent-based model incorporates a fraudster node (F1) attempting to exploit customer nodes within the interbank network.

Strategic Levers and Game Dynamics

The interaction is modeled as a game where players seek Nash equilibria—states where no player can improve their outcome by changing their strategy unilaterally.

  • Bank Strategies: Banks choose a "Bank Threshold" (BT) for RTP eligibility (ranging from 0 to 1,000 units) and an investment level in automated fraud detection (None, Low, or High).
  • Fraudster Strategies: Fraudsters decide whether to target only RTP transactions or to pivot to "Any" payment type, attempting to hide fraudulent activity within standard, slower payment streams.

Key Insights and Results

The Liability Pivot

A critical finding of the study is the impact of fraud liability. When banks carry zero liability for RTP fraud, they typically offer unrestricted access and invest nothing in detection. However, as liability shifts to 100%, banks do not simply disable RTPs. Instead, they strategically layer high-accuracy detection with moderate value thresholds to filter out high-risk transactions.

Fraudster Adaptation and Targeting

Fraudsters are highly responsive to bank defenses. When banks implement robust RTP protections, fraudsters shift their strategy to include standard payments. Furthermore, they tend to "swarm" banks that exhibit historically weak detection or high thresholds, exploiting the path of least resistance.

Strategic Feature Gains

Our Strategic Feature Gains Assessment revealed that setting an initial transaction threshold (BT) is often more critical for initial risk mitigation than the fraud detection software itself. Thresholds serve as a primary filter, while detection software acts as a secondary tool for precision.

Minimal Customer Impact

Encouragingly, the model shows that in equilibrium, bank strategies successfully disrupt fraudsters with negligible impact on legitimate customer transactions. Even when impersonation attempt rates reached up to 80% under stress tests, the actual success rate of fraud remained low due to the effectiveness of layered countermeasures.

Conclusion

As the financial industry moves toward instantaneous settlement, the strategic use of value thresholds and automated detection becomes essential. Our analysis demonstrates that while fraudsters are adaptive, banks can effectively mitigate risk through strategic investment and policy setting, ensuring that the speed of real-time payments does not come at the cost of systemic security.

Summary updated

Abstract

Whereas standard financial mechanisms for payment may take days to finalize, real-time payments (RTPs) provide immediate processing and final receipt of funds. The speed of settlement benefits customers, but raises vulnerability to fraud. We seek to understand how bank nodes may strategically mitigate fraud risk in RTPs, through investment in fraud detection and restricting payments eligible for real-time processing. To study this, we introduce an agent-based model of the payment network supporting both real-time and standard payments, and define a game among banks and fraudsters. Using empirical game-theoretic analysis, we identify Nash equilibria in nine game configurations defined by network attributes. Our analysis finds that as banks become more liable for fraud, they continue to allow RTPs but are more likely to employ both restrictions and a high level of fraud detection. Fraudsters, in response, switch from targeting only RTPs to attempting fraud with any type of payment and tend to exploit banks where they have historically been most successful. We also conduct a strategic feature gains assessment to further understand the benefit offered by each of the bank's risk mitigation measures, which confirms the importance of selective RTP restrictions. Finally, we find that in equilibrium bank strategic decisions negatively affect fraudsters while minimally impacting customers.
BibTeX
@inproceedings{mayo2024fraud,
  author = {Katherine Mayo and Nicholas Grabill and Michael P. Wellman},
  title = {{Fraud Risk Mitigation in Real-Time Payments: A Strategic Agent-Based Analysis}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2024},
  url = {https://openreview.net/forum?id=6tfgADbrgM},
}
Figure