Graph-Derived Features for Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fraud detection methods in telecommunications and internet networking are inadequate in identifying spoofed caller identities and manipulated IP addresses, leading to a need for improved real-time fraud detection mechanisms that leverage the repetitive nature of fraudulent attacks.
Innovation Solution
A computer-implemented system and method using graph-derived features to assess the riskiness of interactions by generating a graph representing structural relationships between communication channels and claimed identities, incorporating metadata to predict the likelihood of fraudulent activity through machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional fraud detection methods are used, then the system is simple to operate, but the detection accuracy is insufficient to identify spoofed caller identities and manipulated IP addresses
Solution Approach 1:
The patent transitions from analyzing individual interaction features to constructing graph structures that capture multi-dimensional relationships between interactions, identities, and communication channels. This graph-based approach adds structural dimensionality to the analysis, enabling detection of complex fraud patterns that conventional methods miss.
Solution Approach 2:
The patent introduces graph structures as an intermediary representation layer between raw interaction data and fraud detection analysis. These graphs serve as mediators that organize and relate multiple data points (interactions, identities, channels) in a structured format, enabling more accurate detection without directly processing the full complexity of raw data.
2Measurement precision
If graph-derived features are used to analyze entire graph structure, then the fraud detection accuracy is improved, but the computational complexity increases
Solution Approach 1:
The patent segments the graph analysis into distinct feature extraction components (node features, edge features, subgraph features) that can be computed and processed separately. This segmentation allows for optimized computation of each feature type and enables parallel processing, reducing overall computational burden while maintaining detection accuracy.
Solution Approach 2:
The patent extracts specific graph-derived features that are most relevant to fraud detection rather than analyzing every possible graph property. This selective feature extraction performs sufficient analysis to achieve high detection accuracy without the excessive computational cost of comprehensive graph analysis.
3Loss of time
If real-time fraud detection is implemented, then the security response time is reduced, but the processing speed requirements increase
Solution Approach 1:
The patent pre-computes and stores graph structures and their derived features during periods when interactions are not being actively evaluated for fraud. This preliminary preparation of graph data and features enables rapid fraud assessment when needed, reducing real-time processing requirements while maintaining fast detection response.
Solution Approach 2:
The patent maintains continuous graph structures that are updated as new interactions occur, rather than rebuilding graphs from scratch for each fraud assessment. This continuous maintenance of graph data allows for efficient incremental updates and rapid fraud detection without repeated full graph construction, improving processing speed.
Data Source
AI summary
Embodiments described herein provide for performing a risk assessment using graph-derived features of a user interaction. A computer receives interaction information and infers information from the interaction based on information provided to the computer by a communication channel used in transmitting the interaction information. The computer may determine a claimed identity of the user associated with the user interaction. The computer may extract features from the inferred identity and claimed identity. The computer generates a graph representing the structural relationship between the communication channels and claimed identities associated with the inferred identity and claimed identity. The computer may extract additional features from the inferred identity and claimed identity using the graph. The computer may apply the features to a machine learning model to generate a risk score indicating the probability of a fraudulent interaction associated with the user interaction.