Anomaly Analysis for Fraudulent Payment Transaction Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic payment systems face challenges in detecting abusive payment transactions, particularly when correct authorization data is used, as they may involve unauthorized access, deception, or intentional misuse, including money laundering, making it difficult to prevent fraud and money laundering effectively.
Innovation Solution
A method using logged transaction data to perform anomaly analysis, identifying anomalies in transaction parameters by comparing probability differences within predefined time intervals, which can indicate abusive payment transactions, allowing for the detection and prevention of such transactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sophisticated authorization checks are implemented to prevent misuse, then security is improved, but false positives increase when correct authorization data is used for fraudulent transactions
Solution Approach 1:
The system segments the detection process into two independent layers: authorization verification (checking if credentials are valid) and anomaly detection (checking if transaction patterns are suspicious). This segmentation allows each layer to specialize - authorization checks validate credentials while anomaly analysis detects fraudulent patterns, resolving the contradiction between security and detection accuracy
Solution Approach 2:
The system introduces an intermediary anomaly analysis layer between authorization verification and transaction execution. This intermediary analyzes transaction parameters (amount, frequency, location, recipient) and compares them against learned patterns to detect fraud even when authorization data is correct, thereby improving detection accuracy without compromising security
2Measurement precision
If anomaly analysis is performed on all transaction parameters, then detection accuracy is improved, but computational complexity increases
Solution Approach 1:
The system applies local quality by focusing anomaly analysis on specific high-risk transaction parameters (amount, frequency, location, recipient) rather than uniformly analyzing all parameters. The anomaly detection mechanism dynamically identifies which parameters show suspicious patterns and concentrates computational resources on those specific aspects, improving detection accuracy while reducing overall complexity
Solution Approach 2:
The system performs partial action by conducting anomaly analysis only on transactions that meet certain risk criteria or show preliminary suspicious indicators. Rather than analyzing every transaction parameter in depth, the system applies anomaly detection selectively to high-priority cases, balancing detection accuracy with computational efficiency
3Speed
If real-time monitoring of transaction parameters is implemented, then fraud detection speed is improved, but processing time increases
Solution Approach 1:
The system performs preliminary action by pre-computing baseline transaction patterns and storing them for quick comparison during live transactions. Statistical models are trained offline on historical data to establish normal behavior patterns, so that during real-time monitoring, the system only needs to compare current transactions against these pre-established baselines, achieving fast detection without excessive processing time
Solution Approach 2:
The system replaces complex real-time computational analysis with simpler statistical comparison mechanisms. Instead of performing heavy machine learning inference on every transaction in real-time, the system uses pre-computed statistical models and simple threshold comparisons, substituting mechanical computation with lighter-weight statistical evaluation to maintain detection speed
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for detecting fraudulent payment transactions (110) using logged transaction data (122) of payment transactions. The method comprises providing logged transaction data (122) of payment transactions, performing an anomaly analysis for the occurrence of fraudulent payment transactions (110), identifying transaction parameters (112) whose values include one or more specific anomalies, and monitoring the values logged for the identified transaction parameters (112) in the transaction data records, wherein the monitoring includes detecting fraudulent payment transactions (110).