Transaction Weighting for Fraud Detection Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fraud detection systems are inadequate for scaling to millions of users in a cost-effective manner, particularly in securing consumer applications, due to their design for lower volume deployments and inability to efficiently handle identity theft and online fraud.
Innovation Solution
A method and system that computes a similarity between transactions by weighting their properties, considering the commonness and distribution of these properties among users and the general population, to improve fraud detection accuracy and prevent fraudulent transactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If enterprise authentication solutions are deployed to secure consumer applications, then security and fraud detection capability are improved, but cost and scalability deteriorate
Solution Approach 1:
The patent transforms enterprise authentication by changing key parameters: shifting from manual review processes to automated machine learning-based transaction scoring, adapting authentication strength dynamically based on risk assessment rather than uniform strong authentication, and modifying the system architecture to handle high-volume consumer transactions cost-effectively
Solution Approach 2:
The system implements dynamic authentication by continuously adjusting authentication requirements based on real-time risk assessment of each transaction. The machine learning model dynamically scores transactions and adjusts the strength of authentication needed, allowing the system to scale efficiently while maintaining security for high-risk transactions
2Reliability
If strong authentication is implemented for all users, then security against identity theft is improved, but user convenience and transaction volume deteriorate
Solution Approach 1:
The patent applies local quality by providing differentiated authentication experiences based on transaction risk characteristics. Low-risk transactions receive minimal authentication requirements maintaining user convenience, while high-risk transactions trigger enhanced authentication measures, ensuring identity protection is applied locally where needed rather than uniformly across all transactions
Solution Approach 2:
The system changes the parameter of authentication strength from a fixed state to a variable state that adjusts based on transaction risk assessment. The machine learning model evaluates transaction properties and dynamically modifies authentication requirements, transforming strong authentication from a blanket approach to a targeted approach that preserves user convenience for low-risk transactions
3Device complexity
If uniform weighting is applied to all transaction properties, then system simplicity is maintained, but fraud detection accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing weightings for numerous transaction properties based on historical fraud patterns and statistical analysis. These pre-calculated weightings are then rapidly applied during live transaction assessment, allowing the system to achieve high detection accuracy without adding operational complexity during transaction processing
Solution Approach 2:
The system transforms the parameter of property weighting from uniform to differentiated by introducing a comprehensive weighting system where each transaction property (device type, location, transaction amount, etc.) has a specific weighting factor. These weightings are derived from historical data and statistical analysis, enabling the system to accurately assess fraud risk while maintaining operational simplicity through automated calculation
Data Source
AI summary
A method of computing a similarity between a first transaction having a set of properties and a second transaction having the set of properties includes computing an initial weight for each of the properties of the set of properties and computing a similarity between each of the properties of the first transaction and the properties of the second transaction. The method also includes adjusting the initial weight for each of the properties based on a measure of the commonness of each of the properties of the set of properties, normalizing the adjusted weights, and computing the similarity by summing the products of the normalized adjusted weights and the computed similarities.