Graph-Based Fraud Detection Using Bipartite Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fraud detection systems in distributed computing environments face challenges in accurately identifying fraudulent behavior due to static rule-based systems and limitations of AI/ML models, which are often biased or inadequate for dynamic fraudster behavior.
Innovation Solution
A graph-based fraud detection system that uses a bipartite graph to represent user accounts and their attributes, allowing for dynamic identification of potentially fraudulent accounts by computing a match score based on shared attributes between monitored and known fraudulent accounts, without requiring training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If rule-based systems are used to identify fraudsters, then the system is simple to implement, but the system is static and easily circumvented by experienced fraudsters who constantly change their behavior patterns
Solution Approach 1:
The patent transforms static rule-based systems into a dynamic graph-based system that continuously evolves. The graph structure dynamically captures relationships between entities (users, devices, locations) and automatically updates when new data is introduced, enabling the system to adapt to changing fraudster behaviors without requiring manual rule updates.
Solution Approach 2:
The system changes the fundamental parameters of fraud detection from fixed rules to dynamic graph representations. By representing fraud detection as a graph problem with nodes and edges, the system can flexibly adjust detection parameters based on the structure and relationships in the graph, rather than being constrained by static threshold-based rules.
2Adaptability or versatility
If AI/ML based models are used to detect fraudulent behavior, then the system can handle dynamic behavior patterns, but the models require a large amount of training data which is not easily available and results in models that are not very accurate, too specific, or biased
Solution Approach 1:
The patent extracts the essential information needed for fraud detection from the complex data and represents it as a graph structure. By taking out key relationships and representing them as edges in the graph, the system eliminates the need for large training datasets while maintaining the ability to detect dynamic behavior patterns through structural analysis.
Solution Approach 2:
The graph structure serves as an intermediary between raw data and fraud detection decisions. Instead of directly training ML models on raw data requiring large datasets, the graph acts as a mediator that captures relationships and patterns, enabling accurate detection without the data requirements of traditional ML approaches.
3Ease of manufacture
If supervised ML models are trained with limited training data, then the system can be implemented with available data, but the resulting models are not very accurate, too specific, or biased
Solution Approach 1:
The patent segments the fraud detection problem into discrete graph components (nodes representing entities, edges representing relationships). This segmentation allows the system to process and analyze data in manageable units, achieving reliable detection results even with limited training data by leveraging the structural relationships captured in the graph segments.
Data Source
AI summary
A fraud detection system is disclosed that detects potential fraudulent behavior associated a user account by identifying attributes of the user account that share attributes with one or more known fraudulent user accounts. The set of shared attributes for a user account are identified by constructing a bipartite graph comprising a set of user account nodes and a set of attribute nodes associated with the set of user account nodes. A match score for the user account is computed based on the set of shared attributes. Actions to be taken for the user account are identified based on the match score. The actions may include tagging the user account as potentially fraudulent. The actions can be used by a user (e.g., an administrator) of an organization to more intelligently determine appropriate measures to be taken for the potentially fraudulent user account.