Fraud Detection via Merchant Transaction Pattern Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fraud detection systems fail to identify test transactions using compromised payment accounts, allowing fraudsters to conduct undetected fraud attacks, leading to increased network load and undetected fraud events.
Innovation Solution
A fraud analysis computing system that utilizes historical transaction data to determine key merchant variables, applies artificial intelligence and machine learning to identify abnormalities in current transaction flows, and flags compromised merchants and test transactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing fraud detection systems monitor payment card transactions for signs of fraudulent activity, then they can detect some fraudulent transactions, but they fail to detect test transactions using compromised payment accounts and allow fraud attacks to proceed undetected
Solution Approach 1:
The system performs preliminary actions by monitoring and analyzing transaction patterns before actual fraud occurs. It detects test transactions (small amount transactions) that use compromised payment accounts before these accounts are used for larger fraudulent purchases. The system establishes baseline merchant transaction patterns and identifies deviations that indicate compromise, enabling early intervention before significant fraud happens.
Solution Approach 2:
The system introduces an intermediary analysis layer between the payment transaction processing and the fraud detection. This intermediary component specifically analyzes merchant transaction patterns, transaction amounts, and timing to identify test transactions. The intermediary system aggregates data from multiple sources including payment card transactions, merchant data, and network patterns to detect compromise indicators that would otherwise be invisible in standard transaction monitoring.
2Reliability
If fraudsters conduct test transactions using compromised payment accounts at merchants, then they can verify account validity before fraud, but the merchants remain unable to detect the hacking
Solution Approach 1:
The system implements feedback mechanisms that continuously monitor merchant transaction patterns and provide real-time alerts when compromise is detected. It compares current transaction data against historical baselines for each merchant and notifies both the payment network and the merchant when abnormal patterns indicate hacking. This feedback loop enables merchants to detect and respond to compromise attempts that would otherwise remain invisible.
Solution Approach 2:
The system replaces traditional mechanical security measures at merchants with network-based pattern recognition and analysis. Instead of relying on merchant systems to detect local anomalies, the system substitutes a centralized intelligence approach that analyzes transaction flows, amounts, and timing patterns across the network to identify compromise without requiring changes to merchant hardware or software.
3Ease of operation
If test transactions are conducted as small amount transactions (e.g., less than $10.00), then fraudsters can verify accounts with minimal risk, but accountholders and issuers generally do not detect the test transactions
Solution Approach 1:
The system changes the detection parameters from traditional transaction amount thresholds to pattern-based anomaly detection. Instead of relying on minimum transaction amounts for fraud detection, the system monitors transaction patterns including frequency, timing, merchant relationships, and behavioral indicators. This parameter change enables detection of small amount test transactions that would be invisible to amount-based monitoring systems.
Solution Approach 2:
The system adds new dimensions to transaction monitoring by analyzing temporal patterns, merchant relationship networks, and behavioral sequences alongside traditional transaction data. It examines the dimension of transaction frequency over time, the dimension of merchant-acccount relationships, and the dimension of transaction pattern consistency to detect test transactions that appear normal in isolation but show anomalous patterns when viewed holistically.
4Measurement precision
If known fraud detection systems score transactions based on transaction characteristics, then they can identify potentially fraudulent transactions, but they require multiple suspicious transactions and are not able to detect test transactions at early stages
Solution Approach 1:
The system performs preliminary analysis by continuously establishing and maintaining baseline transaction patterns for each merchant and account. It monitors for the first signs of compromise (test transactions) before they trigger traditional fraud detection thresholds. This preliminary monitoring enables early detection at the stage of test transactions rather than waiting for multiple suspicious transactions or larger fraud amounts to accumulate.
Data Source
AI summary
Provided herein is a computing system for detecting compromised merchants in a payment card network. The computing system includes a processor in communication with a memory, and the processor is configured to: (i) receive historical transaction data associated with historical transactions carried out at merchants, (ii) determine historical values for key merchant variables for each merchant, (iii) store the historical key merchant variable values, (iv) receive current transaction data associated with current transactions carried out at a selected merchant, (v) determine current values for key merchant variables for the selected merchant, (vi) compare the current key merchant variable values with the historical key merchant variable values for the selected merchant, (vii) identify abnormalities between the current key merchant variable values and the historical key merchant variable values for the selected merchant, and (viii) determine that the selected merchant is a compromised merchant based upon the identified abnormalities.