Behaviorally Validated Link Analysis for Fraud Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for link analysis in fraud detection, such as back-coloring and session linking, often result in a high number of false positives due to reliance on shared resources like IP addresses and geo-locations, leading to incorrect flagging of legitimate transactions and overwhelming investigative efforts.
Innovation Solution
Implementing behaviorally validated link analysis that filters transactions based on unique behavioral characteristics, such as user interactions and device properties, to differentiate between fraudulent and legitimate activities, reducing false positives by verifying criminal behavior signs before expanding the link analysis tree.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If back-coloring and session linking are performed based on shared resources like IP addresses and geo-locations, then the scope of fraud detection is expanded, but the number of false positives increases significantly
Solution Approach 1:
The patent segments the fraud detection process into multiple stages: initial filtering based on behavioral characteristics, intermediate analysis of transaction patterns, and final verification. This segmentation allows the system to progressively narrow down candidates from a large pool of transactions sharing common resources, reducing false positives while maintaining detection scope.
Solution Approach 2:
The patent applies different filtering criteria and analysis methods to different segments of the transaction data. Instead of uniformly flagging all transactions sharing an IP address or geo-location, the system applies localized quality assessments based on behavioral characteristics specific to each transaction context, thereby reducing erroneous flagging while maintaining comprehensive surveillance.
2Reliability
If behavioral characteristics are used to filter transactions, then false positives are reduced, but the complexity of the detection system increases
Solution Approach 1:
The patent performs preliminary analysis of behavioral characteristics during transaction submission or in near-real-time, before the full link analysis is executed. This preliminary action pre-filters transactions based on obvious behavioral red flags, reducing the workload for subsequent complex analysis stages and managing system complexity through staged processing.
Solution Approach 2:
The patent introduces intermediary processing layers that translate raw behavioral data into standardized risk scores or categories. These intermediaries simplify the complexity by providing a structured interface between diverse behavioral inputs and the decision-making logic, making the system more manageable while maintaining high filtering accuracy.
3Reliability
If iterative link analysis is performed to expand the fraud candidate list, then more fraudulent transactions are detected, but legitimate transactions are incorrectly flagged
Solution Approach 1:
The patent implements feedback mechanisms where the results of each iterative link analysis cycle are evaluated against behavioral baselines and historical data. Transactions that emerge as candidates during iterative expansion are continuously re-evaluated, and those lacking consistent behavioral indicators of fraud are filtered out, preventing wrongful incrimination while maintaining detection coverage.
Solution Approach 2:
The patent dynamically adjusts detection parameters and thresholds during iterative link analysis based on the accumulating evidence. As the analysis progresses, the system modifies sensitivity levels, requiring stronger behavioral corroboration for transactions further removed from the seed fraudulent transaction, thereby reducing false positives in later iterations while maintaining detection of strongly correlated fraud.
Data Source
AI summary
System, device, and method for behaviorally validated link analysis, session linking, transaction linking, transaction back-coloring, transaction forward-coloring, fraud detection, and fraud mitigation. A method includes: receiving an indicator of a seed transaction known to be fraudulent; selecting, from a database of transactions, multiple transactions that share at least one common property with the seed transaction; generating a list of candidate fraudulent transactions; filtering the candidate fraudulent transactions, by applying a transaction filtering rule that is based on one or more behavioral characteristics; and generating a filtered list of candidate fraudulent transactions.