Pinpoint Model for Transaction-Level Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fraud detection systems, such as Falcon, primarily focus on identifying cards in a state of fraud rather than distinguishing between legitimate and fraudulent transactions, leading to unnecessary blocking of legitimate transactions during a fraud episode.
Innovation Solution
The pinpoint model is introduced, which uses a cascade model trained on transaction data within a fraud window to differentiate between cardholder and fraudster transactions by employing a Recursive Frequency List (RFL) and adaptive analytics, allowing for the continuation of legitimate transactions and timely reversal of fraudulent ones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional fraud detection systems block all transactions when a card is in a state of fraud, then fraud detection reliability is improved, but legitimate transactions are unnecessarily blocked causing loss of productivity
Solution Approach 1:
The patent segments the fraud detection problem into two distinct levels: card-level fraud detection (Falcon model) and transaction-level fraud detection (pinpoint model). This segmentation allows the system to first identify cards in a fraudulent state, then selectively block only specific fraudulent transactions while allowing legitimate ones to proceed, resolving the contradiction between fraud detection reliability and legitimate transaction throughput.
Solution Approach 2:
The patent applies local quality by transitioning from a uniform approach (blocking all transactions for flagged cards) to a differentiated approach (blocking only fraudulent transactions). The pinpoint model analyzes individual transaction characteristics such as merchant category codes, geographic location, and transaction amount to determine which transactions are fraudulent versus legitimate, allowing legitimate transactions to proceed while blocking only the fraudulent ones.
2Measurement precision
If conventional systems focus on detecting cards in a state of fraud, then fraud detection accuracy is improved, but transaction-level fraud identification capability deteriorates
Solution Approach 1:
The patent segments the analysis into card-level and transaction-level components. The Falcon model handles card-level detection with high accuracy, while the pinpoint model handles transaction-level identification. This segmentation preserves the strengths of card-level detection while adding the capability to identify specific fraudulent transactions, eliminating the loss of transaction-level information.
Solution Approach 2:
The patent introduces the pinpoint model as an intermediary between the Falcon card-level detection and the final transaction blocking decision. The pinpoint model receives the card fraud score from Falcon and then performs detailed transaction-level analysis, acting as a mediator that translates card-level fraud indicators into specific transaction-level fraud identification without losing transaction-level information.
3Loss of energy
If all transactions are blocked when fraud is detected, then fraud loss reversal is optimized, but customer impact increases due to blocking legitimate transactions
Solution Approach 1:
The patent applies local quality by blocking only the specific fraudulent transactions identified by the pinpoint model rather than blocking all transactions. This localized approach minimizes customer impact by allowing legitimate transactions to proceed while still blocking fraudulent ones, thereby optimizing fraud loss reversal without unnecessarily affecting the customer experience.
Data Source
AI summary
A system and method are disclosed, to distinguish fraudulent transactions from a legitimate transaction, predicated on the notion that the card is considered likely to be in state of fraud. The disclosed system and method can be activated as soon as an account has suspicious activity that causes a high score for potential fraud, but before a bank either can or needs to confirm fraud. The system or method is able to pinpoint the actual fraudulent transactions inside a window of potential fraudulent activity, using a specialized model referred to as the pinpoint model.