Identity Fraud Detection via Knowledge Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current identity networks fail to share identity data for discovering fraud correlations and patterns, and they do not share fraud scores, making it difficult to identify fraudulent identities across multiple networks.
Innovation Solution
A method that represents identity profiles as knowledge graphs and uses graph neural networks to associate changes across identity networks with fraud scores, implementing security actions based on these scores to detect and prevent fraudulent activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If identity networks operate independently without sharing data, then each network maintains data security and control, but fraud detection capability across networks deteriorates
Solution Approach 1:
The patent introduces an intermediary system that receives anonymized identity data from multiple networks, processes it through graph neural networks to detect fraud patterns, and returns fraud scores without exposing raw identity data. This mediator enables cross-network fraud detection while preserving data privacy and network autonomy.
Solution Approach 2:
The system extracts only the necessary fraud-relevant features from identity data while removing personally identifiable information. By taking out only the essential patterns needed for fraud detection and leaving the sensitive data behind, the system achieves fraud detection capability without requiring full data sharing.
2Reliability
If fraud scores are not shared across networks, then network security is maintained, but the ability to identify fraudulent identities deteriorates
Solution Approach 1:
Instead of sharing actual fraud scores and sensitive data, the system creates anonymized copies of fraud patterns and correlations. These copied patterns are processed through the graph neural network to generate fraud assessments, allowing accurate identity verification without direct score sharing between networks.
3Reliability
If deep learning models process identity data locally in each network, then processing speed is maintained, but fraud pattern recognition across networks deteriorates
Solution Approach 1:
The patent transitions from local two-dimensional processing within single networks to multi-dimensional cross-network analysis. By organizing identity data from multiple networks into a graph structure with nodes and edges representing relationships, the system enables comprehensive fraud pattern recognition across networks while the distributed architecture maintains processing efficiency.
Data Source
AI summary
A method provides a security action based on identity profile scores. One or more processors represent an identity profile as a knowledge graph. The processor(s) associate a set of changes of the identity profile across a plurality of identity networks with a fraud score. The processor(s) then implement a security action based on the fraud score.