Pay Group Fraud Detection via Group-Level Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fraud detection systems in digital banking are inadequate in identifying and preventing fraudulent activities within pay groups, as they primarily focus on individual transactions rather than group-level correlations, leading to insufficient protection against coordinated attacks on multiple payees.
Innovation Solution
Implementing a multi-step group process that identifies fraudulent activity in a single payment within a pay group, allowing for a group-level decision to flag the entire pay group as fraudulent, and employing flexible fraud prevention rules to restrict enrollments and payments, utilizing advanced pattern recognition, historical transaction data, and machine-learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional fraud detection systems focus on individual transactions, then the system complexity remains manageable, but the ability to detect coordinated attacks on multiple payees is insufficient
Solution Approach 1:
The system segments the pay group into individual payees and transactions for analysis, while also maintaining a group-level view to detect coordinated attacks. This allows the system to handle complexity through structured decomposition while improving fraud detection reliability at both individual and group levels.
Solution Approach 2:
The patent introduces a new dimension of analysis by moving from individual transaction-level detection to group-level pay group detection. This dimensional shift enables the system to detect coordinated fraud patterns across multiple transactions that would be invisible at the individual level, thereby improving reliability without proportionally increasing complexity.
2Reliability
If aggressive fraud prevention policies are implemented to block all suspected fraudulent transactions, then fraud detection reliability improves, but false positives increase causing legitimate users to become upset
Solution Approach 1:
The system applies different quality levels of fraud detection to different payees within a pay group. By analyzing individual payee characteristics and transaction patterns, the system can make localized fraud determinations rather than applying uniform aggressive policies to all transactions, thereby reducing false positives while maintaining high fraud detection accuracy.
Solution Approach 2:
The patent changes the parameters of fraud detection from binary block/approve decisions to a more nuanced risk-assessment model. By adjusting detection parameters based on payee-specific patterns and group-level analysis, the system can reduce false positives while maintaining reliable fraud detection through dynamic parameter adjustment rather than fixed aggressive policies.
3Reliability
If the system analyzes every transaction in detail to reduce false negatives, then fraud detection reliability improves, but the processing time increases
Solution Approach 1:
The system performs preliminary group-level analysis to identify potentially fraudulent pay groups before conducting detailed individual transaction analysis. This preliminary action filters out clearly fraudulent groups early, allowing the system to spend more time on detailed analysis only when necessary, thereby improving sensitivity while managing processing time through staged analysis.
Solution Approach 2:
The patent applies partial analysis to most transactions (standard level) and excessive/detailed analysis only to suspicious transactions that trigger further review. This selective analysis approach improves fraud detection sensitivity for high-risk cases while avoiding the time cost of detailed analysis on all transactions, thereby balancing reliability with processing time.
Data Source
AI summary
Various examples are directed to systems, methods, and computer programs for detecting fraudulent activity and/or fraudulent actors in transaction requests associated with pay groups in a financial institution. For example, a financial server receives a transaction authorization request related to a pay group on an account of a user. The financial server monitors events associated with at least one payee of the pay group and identifies a fraud indication among a first event of the events associated with the at least one payee of the pay group. The financial server blocks the transaction authorization request related to the pay group, including triggering, by the fraud indication, stopping multiple payments to multiple payees of the pay group.