Risk Score Threshold Derivation via Statistical Curve Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional risk-based authentication systems face challenges in updating risk score thresholds effectively, especially when the number of authentication attempts is small, leading to potential false positives or false negatives due to anomalies in risk score distributions.
Innovation Solution
A mathematical estimation operation is performed to derive a risk score threshold by approximating the distribution of risk scores over a time interval using a curve, rather than relying on a specific percentile value, which provides a more robust and accurate representation of authentication policies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a specific percentile value from ranked risk scores is used as the threshold, then the implementation is simple and direct, but the authentication reliability deteriorates when sample size is small due to susceptibility to anomalies
Solution Approach 1:
The patent transforms the threshold selection from a discrete percentile rank approach to a continuous statistical estimation approach. Instead of selecting the top 0.5% risk score directly from ranked data, the system uses mathematical estimation (maximum likelihood estimation) to derive a threshold that corresponds to a specific percentile (e.g., 99.5th percentile) of the underlying risk score distribution. This parameter transformation allows the system to maintain reliability with small sample sizes by estimating the theoretical distribution parameters rather than relying on direct empirical percentiles.
2Adaptability or versatility
If the risk score threshold is updated daily based on past week's data, then the authentication policy remains up to date with current fraud patterns, but the system becomes highly susceptible to anomalies when authentication volumes are low
Solution Approach 1:
The patent implements a feedback mechanism where the risk engine continuously monitors authentication attempts and risk scores, updating the threshold parameters daily based on the observed distribution. The mathematical estimation process uses the past week's risk scores to estimate distribution parameters (mean, standard deviation), and these estimated parameters feed back into the threshold calculation. This feedback loop allows the system to adapt to changing fraud patterns while the statistical estimation provides stability by smoothing out anomalies in low-volume periods.
Data Source
AI summary
A method involves performing a mathematical estimation operation identifying a risk score threshold. The operation identifies the risk score threshold as a point on a curve rather than a value of a particular risk score. Such a curve approximates the distribution of risk score values output over a time interval and represents a function embodied by a plot of risk score percentile vs. risk score value. The risk engine, rather than selecting a particular risk score, selects a curve from a family of curves that is known to accurately represent such risk score distributions. For example, the risk engine may choose the curve that provides the best fit to the previous week's risk scores over the family of curves. The risk engine identifies the risk score threshold by finding a risk score value such that the function evaluated at that risk score value produces a specified risk score percentile.