Collaborative Fraud Detection System Using Multi-Source Data Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for combating credit card and transaction-based fraud are underinclusive, as issuers have limited information to make accurate fraud predictions, while merchants possess relevant data that is not shared with issuers.
Innovation Solution
A method and system for collaborative fraud prevention that involves receiving merchant data, issuer data, third-party metrics, and mobile device data, applying a machine learning model to make fraud predictions, and updating the model with feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If issuers use traditional fraud detection systems with limited data, then the system complexity remains low, but the fraud detection accuracy deteriorates
Solution Approach 1:
The patent merges data from multiple sources including merchant data (device information, purchase history, behavioral patterns), issuer data (account information, transaction history), third-party metrics (credit scores, address verification), and mobile device data (GPS location, accelerometer patterns) into a unified fraud detection system. This combination of previously separate data streams enables comprehensive fraud analysis while maintaining manageable system complexity through integrated processing.
Solution Approach 2:
The patent introduces a communication hub that acts as an intermediary to collect, standardize, and transmit data between merchants, issuers, and third-party services. This intermediary layer simplifies the complexity of direct multi-party data exchange by providing a centralized coordination point that handles data aggregation and distribution according to established protocols.
2Measurement precision
If merchants share detailed transaction data with issuers, then the fraud detection capability improves, but the information security risk increases
Solution Approach 1:
The patent extracts only the specific data elements necessary for fraud detection from the broader set of available merchant data. Instead of sharing all transaction information, the system selectively transmits relevant features such as device identifiers, purchase patterns, and behavioral metrics that directly contribute to fraud prediction while leaving sensitive commercial information at the merchant.
Solution Approach 2:
The patent applies different data sharing protocols to different types of information based on their sensitivity and relevance to fraud detection. Critical fraud-indicative data elements are shared with issuers, while highly sensitive merchant-specific information remains localized. This differentiated approach optimizes fraud detection capability while minimizing security risks associated with broad data exposure.
3Measurement precision
If issuers implement robust fraud prevention systems, then the fraud detection accuracy improves, but the processing time increases
Solution Approach 1:
The patent implements preliminary fraud risk assessment by analyzing merchant data and device information before the actual transaction processing. The communication hub pre-collects and validates data from multiple sources, and the machine learning model pre-evaluates fraud risk indicators, allowing the issuer to make rapid authorization decisions based on pre-processed information rather than analyzing raw data in real-time.
Solution Approach 2:
The patent applies partial processing to transaction data by focusing computational resources on the most fraud-relevant features rather than analyzing every data point in detail. The machine learning model prioritizes analysis of high-risk indicators such as device fingerprint mismatches, unusual purchase patterns, and geographic anomalies, performing deeper analysis only when initial screening indicates potential fraud risk.
Data Source
AI summary
An exemplary method for collaborative fraud prevention comprises receiving, by a processor, merchant data pertaining to a user, receiving, by the processor, issuer data pertaining to the user, receiving, by the processor, third party metrics data pertaining to the user, and receiving, by the processor, mobile device data for a mobile device associated with the user. The exemplary method further comprises applying, by the processor, a machine learning model to the merchant data, issuer data, third-party metrics, and mobile device data to make a fraud prediction, receiving, by the processor, feedback on the fraud prediction, and updating, by the processor, the machine learning model using the feedback as an input.