Device Recognition via MAC Address Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing device identification methods, such as the Hypertext Transfer Protocol (HTTP) standard, result in inconsistent and inaccurate device identification due to varying user-agent strings, making it difficult to manage and secure networked devices effectively.
Innovation Solution
Utilizing a unique identifier like the Media Access Control (MAC) address and supervised machine learning techniques to analyze the remainder of the MAC address, creating a predictive model for accurate device recognition and clustering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If HTTP user-agent strings are used for device identification, then device compatibility is maintained, but identification accuracy deteriorates due to inconsistent and inaccurate device identification
Solution Approach 1:
The patent introduces MAC addresses as an intermediary identifier to bridge the gap between HTTP protocol compatibility and accurate device identification. Instead of relying directly on unreliable user-agent strings, the system uses MAC addresses obtained through HTTP requests as a more reliable intermediate identifier for device recognition and clustering
Solution Approach 2:
The patent replaces the mechanical/string-based user-agent identification system with a data-driven machine learning approach. Supervised learning algorithms are trained on MAC address data to automatically recognize and cluster devices, substituting manual user-agent string parsing with automated pattern recognition
2Measurement precision
If MAC address analysis with machine learning is implemented, then identification precision is improved, but system complexity increases
Solution Approach 1:
The system implements self-service through automated machine learning pipelines that automatically collect MAC address data, train classification models, and deploy recognition algorithms without requiring manual intervention. The supervised learning system self-trains on accumulated device data, continuously improving identification precision while managing its own complexity
Solution Approach 2:
The patent segments the device identification system into distinct modular components: MAC address collection module, data preprocessing module, supervised learning training module, and device recognition deployment module. This segmentation allows each component to be developed and optimized independently, managing overall system complexity while maintaining high identification precision
3Loss of information
If device information is collected from the network, then device recognition capability is improved, but information security risks increase due to potential exposure of sensitive data
Solution Approach 1:
The patent extracts only the necessary MAC address information from network traffic for device identification purposes, leaving sensitive device data on the devices themselves. This selective extraction approach enables device recognition capability while minimizing the collection and storage of sensitive information, thereby reducing information security risks
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments of the present invention provide techniques, systems, and methods for crowdsourcing device recognition to collect device information and identification data from a limited number of network devices and then leverage the collected information with machine learning techniques to expand the starting set in way that the prediction of device attributes like device type, device brand, family and model can be applied on billions of devices.