Multi-Stage Payment Risk Detection With Reinforcement Learning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods and systems for preventing fraudulent electronic payments lack holistic integration across pre-authorization and post-authorization stages, leading to inefficient fraud detection and increased operational latencies, while also causing unnecessary wear on storage components due to excessive data processing.

Innovation Solution

Implementing a policy-based reinforcement learning risk decision agent that utilizes pre-authorization and post-authorization data, along with delay-captured data, to determine the value of impropriety for electronic transactions, thereby enhancing fraud detection and reducing operational latencies by blocking or facilitating transactions based on a threshold value of impropriety.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If current methods process all electronic transmission data through pre-authorization and post-authorization stages, then fraud detection coverage is improved, but operational latency increases and storage component wear increases

Engineering Contradiction:
Improvefraud detection coverageVSAvoidoperational latency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the fraud detection process into distinct stages: pre-authorization stage using a pre-auth model, and post-authorization stage using a chargeback model. This segmentation allows different models to handle different stages independently, reducing overall processing latency while maintaining comprehensive fraud detection coverage across all transaction stages

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements preliminary fraud assessment during the pre-authorization stage before the full transaction processing begins. By performing initial fraud detection early in the transaction flow, the system can identify and block potentially fraudulent transactions before they proceed to later stages, reducing operational latency and preventing storage component wear from processing obviously fraudulent transactions through the entire system

Inventive Principle:
Principle #10Preliminary action

2Reliability

If current methods process all electronic transmission data through multiple authorization stages, then fraud detection accuracy is improved, but storage component wear increases

Engineering Contradiction:
Improvefraud detection accuracyVSAvoidstorage component wear
Core Design Contradiction:
ReliabilityVSObject-generated harmful factors

Solution Approach 1:

The patent extracts and processes only the most relevant features for fraud detection at each stage, rather than processing all available data. The pre-auth model and chargeback model are trained on specific feature sets optimized for their respective stages, reducing the volume of data that needs to be stored and processed while maintaining high fraud detection accuracy

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameters and features processed at different stages of the transaction lifecycle. Different models are trained on different parameter sets appropriate for pre-authorization versus post-authorization contexts, allowing efficient processing with reduced data storage requirements while maintaining detection accuracy through stage-appropriate parameter selection

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4668186A1Multi-point risk detection for electronic transmissions
Publication Date: 2025.12.24 EBAY INC
  • EP4668186A1 patent drawingFigure 1
  • EP4668186A1 patent drawingFigure 2
  • EP4668186A1 patent drawingFigure 3

AI summary

The technology described herein relates to systems, methods, and computer storage media, among other things, for determining whether an electronic transmission (e.g., associated with an electronic payment transaction) should be blocked (e.g., based on being a fraudulent transaction). In embodiments, a policy-based reinforcement learning risk decision agent is used to make these determinations for a plurality of stages associated with the electronic payment transaction (e.g., a pre-authorization stage, a post-authorization stage, and a delay-captured stage). The policy-based reinforcement learning risk decision agent can be trained using previous electronic payment transaction data for previous electronic payment transactions. For example, this particular agent can be trained using pre-authorization electronic payment transaction data, post-authorization electronic payment transaction data, and delay-captured electronic payment transaction data for each of the previous electronic payment transactions.