Detecting Merchant Points of Compromise via Network Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting merchant points of compromise in credit card fraud are not robust or efficient, relying on conventional transaction authorization rules and lacking effective network analysis approaches to identify hidden relationships between fraudulent transactions.
Innovation Solution
A system and method using network analysis and modeling, specifically building undirected and directed transaction networks to extract unique features and create a classification model for automatically detecting suspicious merchant points of compromise, which can be used independently or to enhance existing detection methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional transaction authorization rules are used for fraud detection, then the detection process is simple to implement, but the detection effectiveness and ability to identify hidden relationships is insufficient
Solution Approach 1:
The patent segments the fraud detection problem into multiple network types (undirected transaction network, directed transaction network, account network) and analyzes different aspects of relationships separately. This allows complex fraud patterns to be detected through systematic decomposition into manageable network components, improving detection effectiveness while maintaining organizational clarity.
Solution Approach 2:
The patent transitions from traditional single-dimensional transaction rule-based detection to multi-dimensional network analysis. By creating undirected networks for relationship identification, directed networks for flow analysis, and account networks for pattern recognition, the system adds multiple analytical dimensions that reveal hidden fraud relationships invisible to conventional single-rule approaches.
2Difficulty of detecting and measuring
If network analysis is applied to detect hidden relationships between fraud transactions, then the detection capability improves, but the computational complexity and processing requirements increase
Solution Approach 1:
The patent divides the complex network analysis into separate, manageable network constructions. The undirected transaction network identifies relationship patterns, the directed transaction network analyzes flow directions, and the account network detects compromise patterns. This segmentation reduces computational complexity by processing networks independently rather than analyzing all transactions simultaneously.
Solution Approach 2:
The patent performs preliminary network construction and feature extraction before final fraud determination. By pre-building the networks and extracting meaningful features (degree, betweenness, clustering coefficients), the system prepares data structures that simplify subsequent analysis and reduce real-time computational requirements during fraud detection.
3Productivity
If individual transactions are analyzed separately, then the processing simplicity is maintained, but the efficiency of fraud identification is reduced
Solution Approach 1:
The patent merges multiple transactions into network structures where relationships between transactions become the analysis unit rather than individual transactions. By combining transactions into undirected and directed networks, the system identifies fraud patterns across multiple transactions simultaneously, dramatically improving identification efficiency while the network structure organizes the complexity into manageable forms.
Solution Approach 2:
The patent creates multiple network representations (undirected, directed, account networks) that are copies of the transaction data structured differently. These network copies allow parallel analysis of different relationship types without reprocessing raw transaction data, improving productivity by enabling simultaneous multi-angle fraud detection from prepared network structures.
Data Source
AI summary
Provided is a system for detecting a merchant point of compromise. More specifically, provided is a system for detecting a merchant point of compromise comprising a computer system in electronic communication with a transaction processing network containing transaction information, the computer system comprising a point-of-compromise detector, said point-of-compromise detector performing the steps of electronically receiving from the transaction processing network the transaction information; generating at least one of an undirected network or a directed network based on the transaction information; extracting features from the at least one of the undirected network or the directed network; and identifying one or more point-of-compromise merchants based on the extracted features.