Graph-Based Fraud Detection System for Banking Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods are inadequate in tracing and preventing the rapid dispersion of funds in banking fraud, as they rely on manual investigation and lack real-time identification of fraudulent transactions across multiple bank accounts, making it difficult to recover stolen money and close associated accounts.
Innovation Solution
A system that uses a computer program and apparatus to rapidly identify fraudulent accounts and transactions by building traceable messages through a network of bank accounts, employing graph theory and a Breadth-First traversal method to quickly detect and isolate fraudulent activity, allowing for timely notification and closure of affected accounts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If manual investigation methods are used to trace fraudulent funds, then investigation can be conducted with simple tools, but the tracing speed is too slow and cannot keep up with the rapid transfer of funds
Solution Approach 1:
The patent replaces manual mechanical investigation methods with an automated computer-based system that uses graph theory algorithms to trace fraudulent funds. The system automatically processes transaction data, builds transaction graphs, and identifies fraudulent accounts, eliminating the need for manual tracing while dramatically increasing speed.
Solution Approach 2:
The system performs preliminary actions by continuously monitoring and recording transaction data in real-time, building the transaction graph proactively before fraud occurs. When fraudulent activity is detected, the pre-built graph structure enables immediate tracing without requiring data collection during the investigation, thus accelerating the tracing speed.
2Reliability
If funds are transferred through numerous bank accounts rapidly, then the perpetrator can disperse money and make tracing complicated, but this creates a pattern that can be detected through network analysis
Solution Approach 1:
The patent transitions from traditional linear, bank-by-bank tracing to a multi-dimensional network analysis approach. By representing transactions as a graph with nodes (accounts) and edges (transactions), the system can analyze relationships across multiple dimensions simultaneously, detecting patterns of fund dispersion that would be invisible in traditional sequential analysis.
Solution Approach 2:
The transaction graph system serves multiple functions: it traces individual fraudulent transactions, identifies networks of connected accounts, detects patterns of fund dispersion, and prioritizes investigation targets. This universal approach handles various fraud scenarios uniformly, improving detection reliability across different fraud types.
3Loss of time
If real-time identification of fraudulent transactions is implemented, then money can be stopped and accounts closed quickly, but this requires processing large volumes of transaction data at high speed
Solution Approach 1:
The patent segments the transaction data processing into manageable graph operations. By dividing the transaction network into connected components and using efficient graph traversal algorithms, the system can process large volumes of data incrementally rather than requiring monolithic processing, enabling real-time fraud detection without overwhelming computational resources.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
A method of building a set of traceable messages through a network of nodes is provided, the method comprising receiving message information corresponding to a first message, the message information comprising a first source identifier and a first destination identifier and determining whether the source identifier is associated with a set of previous messages, whereby when the first source identifier is associated with a set of previous messages, the method comprises identifying the first message as also belonging to the set of previous messages and when the first source identifier is not associated with a set of previous messages, the method comprises identifying the first message as belonging to a new set of messages, the method further comprising associating the first destination identifier with the set of messages to which the first message belongs, such that if a second message is sent from that destination, the second message will be identified as belonging to that set of messages.