Grouping IoT Devices via Feature Vectors and Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies lack an effective method to identify and group Internet-connected devices belonging to the same user, especially when these devices use different identifiers and IP addresses, making it difficult for companies to communicate effectively across multiple devices.
Innovation Solution
A system that includes an index structure associating devices with feature information, a pairing engine determining device pairs, a feature vector generation engine producing feature vectors, a scoring engine assigning scores to device pairs, and a clustering engine identifying clusters of devices representing groups of devices associated with the same user, allowing for digital identity grouping and enhanced communication opportunities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If different device identifiers and IP addresses are used for communication, then device anonymity and flexibility are improved, but user identification capability deteriorates
Solution Approach 1:
The patent introduces an intermediary system that collects device feature information (battery level, screen resolution, OS version, etc.) from multiple devices and uses this information as a mediator to infer user identity. Instead of directly identifying users through traditional means, the system uses device feature vectors as intermediate representations to match devices to users based on behavioral patterns and feature similarities.
Solution Approach 2:
The system changes the parameters used for identification from static identifiers (IP addresses, device IDs) to dynamic feature parameters (battery level, screen brightness, OS version, app installation patterns). These parameter changes allow the system to identify users based on their unique device configurations and usage behaviors rather than relying on identifiers that can be easily changed or anonymized.
2Device complexity
If traditional identification methods are used, then user identification is simple, but the ability to track users across multiple devices deteriorates
Solution Approach 1:
The patent adds another dimension to user identification by moving from two-dimensional identification (single device, single identifier) to multi-dimensional identification (multiple devices, multiple feature parameters). The system creates feature vectors that capture device characteristics across multiple dimensions (hardware specs, software versions, usage patterns) enabling reliable cross-device tracking while maintaining a relatively simple overall architecture.
Solution Approach 2:
The identification process is segmented into distinct components: device feature collection, feature vector generation, similarity calculation, and user matching. This segmentation allows the system to handle complex cross-device tracking by breaking it down into manageable steps, where each component can be optimized independently while maintaining overall system simplicity.
3Productivity
If device feature information is collected and analyzed, then user grouping capability is improved, but system complexity increases
Solution Approach 1:
The system creates simplified copies of device information in the form of feature vectors. Instead of collecting and processing all raw device data, the system extracts essential features and creates condensed vector representations that capture the essential characteristics needed for user identification. This copying approach improves grouping efficiency while avoiding the complexity of processing complete device information sets.
Solution Approach 2:
The system performs preliminary actions by pre-collecting and pre-processing device feature information before user grouping is needed. Device features are continuously monitored and stored in advance, and feature vectors are pre-computed so that when user grouping is required, the system can quickly perform similarity calculations without the complexity of real-time feature extraction and analysis.
Data Source
AI summary
A system comprising: an index structure that associates devices with device feature information; a pairing engine to determine device pairs based upon device feature information; a feature vector generation engine to produce feature vectors corresponding to determined device pairs based upon feature values associated within the index structure with devices of the determined device pairs; a scoring engine to determine scores to associate with determined device pairs based upon produced feature vectors; a graph structure, wherein nodes within the graph structure represent devices of determined device pairs, and wherein edges between pairs of nodes within the graph structure indicate determined device pairs; a clustering engine to identify respective clusters of three or more nodes within the graph structure that represent respective groups of devices.