User Effort Pattern Analysis for Reliable Anomaly Authentication
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user authentication methods using statistical analysis of human effort and movement metrics face challenges in accurately identifying and removing anomalous data due to unknown variability, as they rely on normal distribution models that do not apply to human input, leading to unreliable results.
Innovation Solution
A system that captures and analyzes the cadence and habit of users' motions during input on devices like keyboards, forming a PCCHL, and uses Euclidean space analysis to compare current inputs with stored profiles, employing geometric and geospatial constructs to authenticate users by measuring gravitational-like forces between input patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional normal distribution models and known reference standards are used for outlier detection, then the analysis process is simple and familiar, but the accuracy and reliability of anomaly identification deteriorate because these models do not apply to human input variability
Solution Approach 1:
The patent transforms human effort and movement metrics into statistical parameters (mean, standard deviation, coefficient of variation) that characterize normal human variability. By changing the parameters from fixed reference standards to dynamic statistical descriptors of human behavior, the system achieves accurate anomaly detection that accounts for inherent human variability without requiring complex machine learning models.
Solution Approach 2:
The patent introduces statistical analysis as an intermediary between raw human input data and anomaly detection. Instead of directly comparing human input to rigid reference standards, the system uses statistical parameters as a mediator to normalize and characterize human variability, enabling accurate anomaly identification while maintaining analytical simplicity.
2Reliability
If machine learning and artificial intelligence technologies are used to resolve anomalous information identification, then the accuracy of anomaly detection improves, but the device complexity and computational requirements increase significantly
Solution Approach 1:
The patent extracts only the essential statistical characteristics (mean, standard deviation, coefficient of variation) from human input data that are necessary for anomaly detection. By taking out only these critical parameters rather than using comprehensive machine learning models, the system achieves reliable anomaly detection with minimal computational complexity and without requiring AI technologies.
Solution Approach 2:
The patent uses simple, computationally inexpensive statistical calculations instead of expensive and complex machine learning models. These statistical parameters can be quickly computed and discarded after use, providing reliable anomaly detection without the high computational cost and complexity of AI systems.
3Ease of operation
If known reference standards are used to evaluate human input data, then the evaluation process is straightforward, but the reliability deteriorates because human variability does not conform to these fixed standards
Solution Approach 1:
The patent transitions from static known reference standards to dynamic statistical parameters that adapt to each user's normal variability. By making the evaluation criteria dynamic rather than fixed, the system maintains ease of operation while significantly improving authentication reliability to account for inherent human variability in effort and movement metrics.
Data Source
AI summary
A variety of systems and methods can include evaluation of human user effort data. Various embodiments apply techniques to identify anomalous effort data for the purpose of detecting the efforts of a single person, as well as to segment and isolate multiple persons from a single collection of data. Additional embodiments describe the methods for using real-time anomaly detection systems that provide indicators for scoring effort data in synthesized risk analysis. Other embodiments include approaches to distinguish anomalous effort data when the abnormalities are known to be produced by a single entity, as might be applied to medical research and enhance sentiment analysis, as well as detecting the presence of a single person's effort data among multiple collections, as might be applied to fraud analysis and insider threat investigations. Embodiments include techniques for analyzing the effects of adding and removing detected anomalies from a given collection on subsequent analysis.