AML Threat Score Using Behavior-Sorted Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anti-money laundering (AML) systems face challenges in effectively detecting and prioritizing suspicious transactions due to the unbalanced nature of training data and high false positives, which limits their ability to accurately identify money laundering activities.
Innovation Solution
An automated system using behavior-sorted lists and threat scores is developed to efficiently summarize entity behavior, prioritize alerts, and link entities, incorporating self-calibrating outlier models and recursive features to generate an AML Threat Score that indicates the likelihood of money laundering activity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional rules-based systems are used to meet regulatory requirements, then compliance with regulations is achieved, but the systems are easily circumvented by breaking up large transactions into smaller amounts
Solution Approach 1:
The patent transitions from static rules-based systems to dynamic machine learning models that adapt to new laundering patterns. The system continuously learns from transaction data and updates its detection capabilities, making it resistant to circumvention techniques that static rules cannot anticipate.
Solution Approach 2:
The system changes the parameters of detection by moving from fixed threshold rules to probabilistic models that consider multiple transaction attributes simultaneously. This allows the system to detect laundering patterns that don't fit predefined rules by analyzing combinations of parameters like transaction frequency, amount patterns, and counterparty relationships.
2Measurement precision
If supervised machine learning methods are used with unbalanced training data, then detection accuracy for known patterns improves, but the highly unbalanced nature of SAR filings limits effectiveness
Solution Approach 1:
The system performs preliminary analysis by generating alerts for potentially suspicious transactions before final SAR determination. This allows the machine learning model to practice on a larger set of candidate cases, effectively creating a balanced training dataset where both positive and negative examples are represented, improving overall detection accuracy.
Solution Approach 2:
The patent introduces an intermediary alerting layer between transaction monitoring and SAR filing. This intermediary system generates alerts that can be used for training, creating a balanced dataset by including both alerted and non-alerted cases, thereby improving the machine learning model's ability to distinguish true laundering patterns from normal variations.
3Productivity
If traditional AML systems generate high-volume alerts, then comprehensive monitoring is achieved, but only a small fraction of alerts lead to SAR filings, indicating high false positives
Solution Approach 1:
The patent segments the alert generation process into multiple stages with different machine learning models specializing in different aspects of laundering detection. This segmentation allows each model to focus on specific patterns, reducing false positives while maintaining comprehensive monitoring coverage across different transaction types and laundering techniques.
Solution Approach 2:
The system implements feedback loops where alert outcomes (whether they lead to SAR filings or not) are used to retrain and improve the machine learning models. This continuous feedback reduces false positives over time by learning from actual SAR decisions, while maintaining high alert volumes for comprehensive monitoring of emerging patterns.
Data Source
AI summary
An automated system for detecting risky entity behavior using an efficient frequent behavior-sorted list is disclosed. From these lists, fingerprints and distance measures can be constructed to enable comparison to known risky entities. The lists also facilitate efficient linking of entities to each other, such that risk information propagates through entity associations. These behavior sorted lists, in combination with other profiling techniques, which efficiently summarize information about the entity within a data store, can be used to create threat scores. These threat scores may be applied within the context of anti-money laundering (AML) and retail banking fraud detection systems. A particular instantiation of these scores elaborated here is the AML Threat Score, which is trained to identify behavior for a banking customer that is suspicious and indicates high likelihood of money laundering activity.