Fraud Ring Detection via User Clustering and Aggregated Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fraud rings, consisting of organized and methodical fraudulent users, pose a significant threat by masking their activities effectively, making it difficult for existing technologies to distinguish them from legitimate users, leading to potential greater damage compared to individual fraudulent actors.
Innovation Solution
A method and system for detecting fraud rings by clustering users based on grouping and homogeneity attributes, determining aggregated features, calculating predictive or label-based suspiciousness scores, and taking protective actions against suspicious clusters, utilizing machine learning classifiers and clustering techniques to identify and flag potentially fraudulent user groups.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional fraud detection methods are used to identify individual fraudulent users, then individual fraud cases can be detected, but organized fraud rings can mask their activities and evade detection
Solution Approach 1:
The patent segments the fraud detection problem from individual user analysis to cluster-based analysis. Users are grouped into clusters based on homogeneity attributes (similar behaviors, demographics, device characteristics), and fraud detection is performed at the cluster level rather than individually. This segmentation enables the system to detect organized fraud rings as cohesive groups while maintaining manageable complexity through standardized clustering algorithms and aggregated feature analysis.
Solution Approach 2:
The patent merges multiple user accounts into unified clusters based on shared attributes, treating them as a single analytical unit. By combining individual user data points (behaviors, device info, transaction patterns) into cluster-level aggregated features, the system achieves more robust fraud detection that captures the coordinated nature of fraud rings while reducing the complexity of analyzing each account separately.
2Reliability
If multiple user accounts are analyzed individually for fraud, then each account can be assessed separately, but the coordinated nature of fraud rings is missed
Solution Approach 1:
The patent segments users into homogeneous clusters based on shared attributes, transforming the analysis from individual account level to cluster level. This segmentation improves reliability by capturing the coordinated patterns of fraud rings while maintaining productivity through efficient aggregate computations on cluster-level features rather than processing each account independently.
Solution Approach 2:
The patent creates a universal clustering framework that can process multiple user accounts simultaneously using standardized homogeneity attributes and aggregation methods. This multi-functional approach enables the system to analyze diverse fraud ring compositions (different industries, account types, fraud methods) through a single unified process, improving both reliability across different fraud scenarios and productivity through batch processing capabilities.
3Measurement precision
If fraud detection systems analyze detailed individual user behaviors, then accurate fraud identification is possible, but computational resources and processing time increase significantly
Solution Approach 1:
The patent merges individual user behavior data into cluster-level aggregated features, combining multiple detailed observations into summarized statistics (means, variances, distributions). This merging maintains detection precision by preserving essential fraud patterns in aggregate form while dramatically reducing computational resource consumption compared to analyzing every individual user behavior in detail.
Solution Approach 2:
The patent transforms detailed individual behavior parameters into aggregated cluster-level parameters through statistical transformations. By changing from raw individual data parameters to derived aggregate parameters (distributions, correlations, homogeneity metrics), the system maintains the ability to detect fraud patterns while reducing the dimensional complexity and computational burden of the analysis.
Data Source
AI summary
A method for detecting fraud rings involves clustering unknown users into unknown user clusters based on a grouping attribute. The method further involves, for each of the unknown user clusters, determining aggregated features including at least one quantification of at least one homogeneity attribute across the unknown users in the unknown user cluster. The method also involves, for each of the unknown user clusters, determining a predictive suspiciousness score based on the aggregated features, determining that at least one of the unknown user clusters is suspicious based on the determined predictive suspiciousness scores, and taking a protective action for the at least one suspicious unknown user cluster.