Multivariate Outlier Detection for Healthcare Data Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data anonymization techniques struggle to effectively detect and address multivariate outliers in high-dimensional datasets, which can compromise data privacy and are difficult to identify using conventional methods.
Innovation Solution
A computer-implemented method for data protection that performs multivariate outlier detection using algorithms like isolation forest, elliptic envelope, and local outlier factors, computing anomaly scores to rank datapoints and de-identify sensitive information, with optional user input for algorithm selection and explainable AI for insight into outlier scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If conventional univariate outlier detection methods are used, then detection simplicity is maintained, but detection precision for multivariate outliers deteriorates
Solution Approach 1:
The patent transitions from univariate to multivariate outlier detection by computing anomaly scores that consider multiple variables simultaneously. The system evaluates data points in a multidimensional space rather than analyzing single variables in isolation, enabling detection of outliers that only become apparent when multiple attributes are combined.
Solution Approach 2:
The patent combines multiple outlier detection algorithms (isolation forest, elliptic envelope, local outlier factors) into a composite detection system. This ensemble approach integrates the strengths of different algorithms to achieve higher detection precision while maintaining operational simplicity through automated score aggregation and ranking.
2Productivity
If data anonymization is performed without multivariate outlier detection, then data utility is preserved, but data privacy protection deteriorates
Solution Approach 1:
The patent applies multivariate outlier detection before data anonymization to identify and flag suspicious data points that could enable re-identification. By detecting anomalies in advance, the system allows targeted protection measures to be applied only to high-risk data points, preserving the utility of normal data while enhancing privacy protection where needed.
Solution Approach 2:
The patent applies different protection strategies to different data points based on their anomaly scores. High-scoring outliers receive enhanced anonymization or removal, while low-scoring data points maintain their original quality. This localized approach preserves data utility for the majority of data while providing targeted privacy protection for suspicious cases.
3Measurement precision
If multiple multivariate outlier detection algorithms are employed, then detection precision is improved, but device complexity increases
Solution Approach 1:
The patent merges multiple outlier detection algorithms into a unified system that computes anomaly scores from each algorithm and aggregates them. The isolation forest, elliptic envelope, and local outlier factor algorithms work together in parallel, with their results combined through a scoring mechanism that ranks data points by overall anomaly level, achieving high precision without requiring complex sequential processing.
Solution Approach 2:
The patent creates a universal outlier detection framework that can accommodate multiple different algorithms through a common interface. The system evaluates various algorithms (isolation forest, elliptic envelope, local outlier factors) using the same computational structure and score aggregation method, making the system flexible and adaptable while maintaining manageable complexity through standardized processing.
Data Source
AI summary
A computer-implemented data protection method comprising: receiving an input dataset, the input dataset including a plurality of datapoints, at least some of the plurality of datapoints including information usable in combination to identify a patient; performing multivariate outlier detection on the input dataset, the performing including computing anomaly scores for at least a portion of the plurality of datapoints using a multivariate outlier detection algorithm; and identifying, based on the anomaly scores, at least one set of multivariate outliers of datapoints usable in combination to identify the patient.