Graph Embedding for Robocalling Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for telecommunication network fraud and threat detection face challenges in efficiently processing large volumes of call detail records and identifying complex patterns like robocalling behaviors, leading to high processing costs and time consumption.
Innovation Solution
The method involves constructing a communication graph with both long-term and short-term perspectives, using graph embedding processes to generate vectors that are applied to prediction models for real-time or near-real-time detection of robocalling and fraud activities, thereby reducing computational costs and improving detection speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used to process call detail records for fraud detection, then detection accuracy can be maintained, but processing time and computational costs increase significantly
Solution Approach 1:
The patent segments the telecommunication network data into graph structures where phone numbers are nodes and calls are edges. This segmentation allows the system to process and analyze fraud patterns more efficiently by working with structured graph data rather than raw call detail records, reducing processing time while maintaining detection accuracy through graph embedding techniques.
Solution Approach 2:
The patent replaces traditional mechanical processing methods of call detail records with graph-based computational approaches. By using graph embeddings and machine learning models on graph structures, the system achieves faster processing speeds while maintaining or improving fraud detection accuracy compared to conventional methods.
2Reliability
If comprehensive analysis of communication patterns is performed to identify fraud rings, then detection reliability improves, but computational complexity increases
Solution Approach 1:
The patent transforms communication pattern data into graph structures with multiple dimensions - nodes representing phone numbers, edges representing calls, and additional attributes capturing temporal and relational information. This dimensional transformation enables comprehensive fraud ring detection while managing computational complexity through efficient graph algorithms and embedding techniques.
Solution Approach 2:
The patent changes the parameters of data representation from raw call detail records to graph-based features including node degrees, clustering coefficients, and graph embeddings. These parameter transformations enable more efficient computation of fraud patterns while maintaining or improving detection reliability through richer feature representation.
3Reliability
If real-time fraud detection is implemented, then network protection effectiveness improves, but processing speed requirements increase computational demands
Solution Approach 1:
The patent performs preliminary actions by pre-computing graph embeddings and structural features from communication patterns. These pre-computed representations are stored and can be quickly queried during real-time fraud detection, reducing the computational burden during actual detection operations while maintaining high protection effectiveness.
Solution Approach 2:
The patent creates compressed representations (graph embeddings) that copy the essential structural and relational information of the communication graph in a more compact form. These embeddings enable real-time fraud detection with reduced computational resource consumption while preserving the ability to detect complex fraud patterns.
Data Source
AI summary
A processing system may maintain a communication graph that includes nodes representing a plurality of phone numbers including a first phone number and edges between the nodes representing a plurality of communications between the plurality of phone numbers and may generate at least one vector via a graph embedding process applied to the communication graph, the at least one vector representing features of at least a portion of the communication graph. The processing system may then apply the at least one vector to a prediction model that is implemented by the processing system and that is configured to predict whether the first phone number is associated with a type of network activity associated with a telecommunication network and may implement a remedial action in response to an output of the prediction model indicating that the first phone number is associated with the type of network activity.