POS Fraud Detection via Local Cluster Matching

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current fraud detection systems for card-based transactions at POS devices are inefficient and prone to errors, often requiring communication with remote entities and relying on network bandwidth, which can lead to delays and increased exposure to fraudulent activities.

Innovation Solution

Implementing a POS device that obtains sensory inputs and identifiers from cards and mobile devices, maps these inputs to clusters of learned features using machine learning, and initiates second-factor authentication when the transaction does not match expected behavior, allowing for real-time fraud detection and reduced reliance on remote systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If fraud detection systems communicate with remote entities for transaction verification, then authentication accuracy can be improved, but network bandwidth dependency increases and transaction delays occur

Engineering Contradiction:

Improvefraud detection accuracyVSAvoidtransaction processing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The fraud detection system is segmented into local and remote components. The POS device performs local sensory input collection and cluster matching, while remote entities handle complex fraud analysis. This segmentation allows immediate local processing to reduce delays while remote verification maintains high accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Sensory inputs are collected and clustered in advance at the POS device before transaction completion. By performing preliminary local analysis and only contacting remote entities when needed, the system reduces network dependency and accelerates routine transactions while maintaining security.

Inventive Principle:

Principle #10Preliminary action

2Device complexity

If traditional fraud detection methods are used, then system simplicity is maintained, but fraud detection effectiveness decreases

Engineering Contradiction:

Improvesystem structureVSAvoidfraud detection effectiveness

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

Machine learning clusters serve as an intermediary between simple sensory inputs and fraud detection decisions. These pre-computed behavioral patterns enable the POS device to perform sophisticated fraud analysis locally without requiring complex real-time computation, bridging simplicity and effectiveness.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

Traditional rule-based fraud detection is replaced with machine learning-based cluster matching. This substitution enables more accurate behavioral pattern recognition while maintaining system simplicity through pre-computed clusters stored locally at the POS device.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If multiple authentication factors are always required, then security is improved, but transaction processing speed decreases

Engineering Contradiction:

Improvetransaction securityVSAvoidtransaction processing throughput

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The system applies partial authentication based on risk assessment. For low-risk transactions matching expected clusters, minimal authentication is performed. For suspicious transactions, full multi-factor authentication is triggered. This partial approach maintains security while optimizing processing throughput.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The authentication requirement dynamically adjusts based on transaction characteristics and cluster matching results. The system transitions between simple and complex authentication modes, optimizing the balance between security and processing speed for each individual transaction.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS20240420148A1Computer-based systems having computing devices programmed to execute fraud detection routines based on feature sets associated with input from physical cards and methods of use thereof

Publication Date: 2024.12.19 CAPITAL ONE SERVICES LLC

AI summary

Systems and methods for performing fraud detection at POA devices based on analysis of feature sets are disclosed. In one embodiment, an exemplary method may comprise: obtaining, by a POS device, upon initiation of a transaction involving a card or a card and mobile device associated with an individual initiating the transaction, one or more sensory inputs and an identifier; mapping, by the POS device, the one or more sensory inputs to a first cluster position of a plurality of clusters; determining whether the cluster position of the cluster mapped for the transaction corresponds to a second cluster position of the at least one expected cluster associated with the known owner of the card and/or mobile device; and initiating, by the POS device, at least one second factor authentication process to establish that the individual is the known owner of the card and/or mobile device being used in the transaction.