Fraud Risk Mitigation in Real-Time Payments: A Strategic Agent-Based Analysis
Authors: Katherine Mayo, Nicholas Grabill, Michael P. Wellman
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 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. 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. |
| Researcher Affiliation | Academia | Katherine Mayo , Nicholas Grabill and Michael P. Wellman University of Michigan {kamayo, grabilln, wellman}@umich.edu |
| Pseudocode | No | The paper describes the model and game mechanics in detail but does not include any pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any specific links or statements regarding the public availability of its source code. |
| Open Datasets | No | The paper describes initializing an agent-based model with customer and bank nodes and their properties (e.g., "We initialize n = 200 customer nodes with initial deposits drawn from an exponential distribution with an average value of 10,000."). It does not use or provide access to a pre-existing, publicly available dataset in the traditional sense for training. |
| Dataset Splits | No | The paper describes its simulation methodology, including running "3,000 random generations of the game" and analyzing outcomes "over 1,000 runs of the game" for EGTA. However, it does not specify traditional train/validation/test dataset splits as it uses a simulation-based approach rather than a fixed dataset. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used to run the experiments or simulations. |
| Software Dependencies | No | The paper does not provide specific version numbers for any software libraries, platforms, or tools used in their implementation or experiments. |
| Experiment Setup | Yes | The RTPs fraud game unfolds over T = 2, 880 time steps with each time step containing payment creation, fraud detection, and possible network updates. Every 96 time steps (the clearing period), the bank queues are cleared... During the payment creation phase, η = 20 customer nodes are randomly selected... With probability µ = 0.7 a fraudster node is one of the η selected. The value of the payment is drawn from v = U (0, min(1000, d)]. We initialize n = 200 customer nodes... Each customer node is also initialized with its own personal threshold, CT i {200, 400, 600, 800, 1000}... We model this with the urgency parameter such that with probability u = 0.5... A set of b = 4 banks nodes is initialized... Bank node Bi s standard payments fraud detector γS i is initialized with a random draw from U [0.8, 0.9]... The strategy set for bank nodes contains all 24 combinations of threshold choices and investment levels. |