Mobile User Authentication with Machine-Learned Behavior Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing user authentication methods on mobile devices, such as passcodes and biometrics, are vulnerable to circumvention and theft, posing a security risk as smartphones become gateways to sensitive systems and IoT devices.
Innovation Solution
Implement machine-learning based user authentication using motion and input patterns on mobile devices, utilizing sensors like accelerometers and pressure sensors to recognize authorized user habits and patterns, and employ alternate authentication methods such as video and audio comparisons when primary authentication fails.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional passcodes are used for authentication, then ease of operation is maintained, but security reliability deteriorates as passcodes can be circumvented or stolen
Solution Approach 1:
The patent replaces traditional mechanical/passcode-based authentication with machine learning-based behavioral biometric authentication. The system uses sensors to collect motion data and analyzes user behavior patterns through machine learning algorithms, substituting the manual passcode entry mechanism with automated behavioral analysis that occurs transparently in the background.
Solution Approach 2:
The authentication system performs self-service by automatically collecting behavioral data through device sensors, continuously learning user patterns through machine learning, and autonomously making authentication decisions without requiring active user participation. The system serves itself by using its own operational data to improve authentication accuracy over time.
2Reliability
If machine learning based authentication is implemented, then authentication security is improved, but device complexity increases due to additional sensors and processing requirements
Solution Approach 1:
The patent leverages existing multi-functional sensors in modern mobile devices (accelerometers, gyroscopes, touch screens, cameras) that were originally designed for other purposes like gaming, navigation, and user interface interaction. By repurposing these universal components for authentication, the system avoids adding dedicated hardware while achieving enhanced security.
Solution Approach 2:
The patent introduces machine learning algorithms as an intermediary layer that processes raw sensor data and translates it into authentication decisions. This intermediary component bridges the gap between simple sensor inputs and complex security requirements, managing system complexity through intelligent software rather than hardware complexity.
3Measurement precision
If behavioral patterns are analyzed for authentication, then measurement precision of user identification is improved, but loss of information occurs due to the need to process and store motion data
Solution Approach 1:
The system performs preliminary actions by continuously collecting and pre-processing behavioral data in the background during normal device usage. Machine learning models are trained offline on accumulated data, so when authentication is needed, the system can quickly compare current behavior against pre-established patterns without requiring extensive real-time data processing or storage.
Solution Approach 2:
The patent segments the authentication process into distinct phases: data collection through sensors, feature extraction from raw data, pattern matching against learned behaviors, and final authentication decision. This segmentation allows the system to process information in manageable chunks, storing only essential pattern representations rather than all raw data, thereby reducing information loss and storage requirements.
Data Source
AI summary
Machine-learning based user authentication using a mobile device (e.g., using a computerized tool) is enabled. For example, a non-transitory machine-readable medium can comprise executable instructions that, when executed by a processor, facilitate performance of operations, comprising: determining an input received via a mobile device, determining, based on the input and using an authentication model, whether the input threshold matches an input pattern associated with an authorized user profile authorized to access a feature of the mobile device, wherein the input pattern has been determined based on machine learning applied to past inputs at the mobile device other than the input, and wherein the authentication model has been generated based on the machine learning applied to the input pattern, and based on a determination that the input at the mobile device is associated with an authorized user profile, granting access to the feature of the mobile device.