Transaction Flow Graph for Collaborative Fraud Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Traditional methods for detecting abnormal transactions in electronic payment processing networks often fail to identify complex security attacks, particularly collaborative frauds, as they treat individual transactions separately and struggle to connect transactions with shared financial or location attributes.

Innovation Solution

A system utilizing a transaction flow graph that connects transactions hierarchically, using two trees representing financial and location hierarchies, to aggregate small fraud values and detect abnormal activity in real-time without pre-training a model, enabling the identification of collaborative frauds.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional supervised and unsupervised models are used to detect abnormal transactions, then individual transactions can be processed, but collaborative frauds involving multiple accounts cannot be detected

Engineering Contradiction:

Improvedetection accuracyVSAvoidability to detect collaborative frauds

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent combines multiple individual transaction analyses into a unified graph-based system that detects collaborative frauds. By merging transactions into a transaction flow graph where nodes represent accounts and edges represent transactions, the system can identify fraud patterns across multiple accounts that individual transaction analysis cannot detect.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent transitions from analyzing transactions in one dimension (individual transaction attributes) to multiple dimensions by constructing a graph structure that adds relational dimensions. The graph framework incorporates financial hierarchy, location hierarchy, and temporal dimensions, enabling detection of complex fraud patterns that require multi-dimensional analysis.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Device complexity

If transactions are treated separately in traditional models, then processing is simpler, but abnormal patterns across multiple transactions cannot be identified

Engineering Contradiction:

Improvemodel complexityVSAvoidabnormal pattern detection

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent segments the complex fraud detection problem into hierarchical components: financial hierarchy (accounts, merchants, locations), location hierarchy (countries, cities, regions), and temporal hierarchy. This segmentation allows the system to manage complexity by breaking down the overall detection task into manageable hierarchical levels while maintaining the ability to detect cross-level patterns.

Inventive Principle:

Principle #1Segmentation

3Reliability

If hierarchical transaction flow graphs are used to detect collaborative frauds, then detection accuracy improves, but real-time processing becomes more challenging

Engineering Contradiction:

Improvecollaborative fraud detectionVSAvoidreal-time processing speed

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent performs preliminary actions by pre-establishing the hierarchical graph structure and relationships between accounts, merchants, and locations before fraud detection is needed. The transaction flow graph is constructed with predefined financial and location hierarchies, allowing real-time fraud detection to focus on analyzing incoming transactions against the pre-built structure rather than building the structure itself in real-time.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11551230B2Security attack detections for transactions in electronic payment processing networks

Publication Date: 2023.01.10 VISA INTERNATIONAL SERVICE ASSOCIATION

AI summary

Systems, apparatuses, methods, and computer-readable media are provided for detecting security attacks based on transaction flow graphs. Other embodiments may be described and/or claimed.