Message Motif Construction for Fraud Detection in Banking Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity and high transaction rates in modern computer networks, particularly in banking systems, make it computationally demanding to distinguish between legitimate and fraudulent financial transactions, leading to delays in identifying fraudulent activities.
Innovation Solution
A method and apparatus for constructing motifs from target and control data to identify patterns of messages of interest, using motif detection and significance analysis to differentiate between legitimate and fraudulent transactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual tracing methods are used to investigate banking fraud, then investigation accuracy can be maintained, but the time required to trace funds increases significantly
Solution Approach 1:
The patent replaces manual mechanical investigation methods with an automated computational system that uses motif detection algorithms to automatically trace fraudulent transactions through banking networks, achieving both speed and accuracy simultaneously
Solution Approach 2:
The system creates simplified copies or representations of complex transaction patterns through motifs, which are condensed models of message sequences that can be rapidly analyzed to trace fraud without processing every individual transaction in detail
2Productivity
If the banking network processes a high volume of transactions at high speed, then productivity is improved, but the complexity of distinguishing legitimate from fraudulent transactions increases
Solution Approach 1:
The patent extracts only the essential and distinguishing features of transaction patterns by identifying motifs, which are representative sequences of messages that capture the essence of fraudulent behavior without requiring analysis of the entire complex transaction dataset
Solution Approach 2:
The system changes the parameters of analysis by shifting from examining individual transaction details to detecting recurring message patterns (motifs), thereby simplifying the detection process while maintaining effectiveness in high-volume environments
3Measurement precision
If comprehensive analysis of all messages is performed to ensure accurate fraud detection, then detection precision is improved, but computational demand increases
Solution Approach 1:
The patent uses motifs as simplified copies or representations of complex message sequences, allowing the system to analyze pattern representations rather than every individual message, thereby reducing computational demand while preserving detection accuracy
Solution Approach 2:
The system performs partial analysis by focusing only on detecting specific recurring patterns (motifs) rather than comprehensively analyzing all message attributes, achieving sufficient detection accuracy with reduced computational effort
Data Source
AI summary
A method of constructing a set of motifs for use in detecting messages of interest in a network of nodes is provided, the method comprising controlling circuitry to: acquire target data, the target data comprising a set of messages which have been exchanged between nodes in the network, the set of messages including a number of messages of interest; acquire control data, the control data comprising a set of messages which have been produced based on a random exchange of messages between nodes in the network; detect motifs within the target data and the control data, each motif being a repeated pattern of messages appearing within either the target data and/or the control data; generate a set of values indicative of a significance of the motifs which have been detected in the target data and the motifs which have been detected in the control data using a frequency with which these motifs have been detected; and construct a set of motifs for use in detecting messages of interest in the network using the set of values which have been produced and a comparison of the motifs detected in the target data with the motifs detected in the control data.