Device Fingerprinting via Neural Network Sensor Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing device fingerprinting methods are vulnerable to spoofing, as attackers can modify user-agent and hardware information, making it difficult to reliably identify computing devices and prevent fraud.
Innovation Solution
A system that generates device fingerprints using sensor data from accelerometers, gyroscopes, and other sensors, combined with time-domain and frequency-domain features, and processed through a neural network to create a unique fingerprint, which is then validated and fused with additional device information to enhance identification reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If device fingerprinting is based on user-agent and hardware information, then device identification can be performed, but the system becomes vulnerable to spoofing and manipulation
Solution Approach 1:
The patent replaces software-based fingerprinting (user-agent, hardware info) with physics-based sensor data collection. Sensors like accelerometers, gyroscopes, and ambient light sensors capture physical characteristics that are difficult to spoof programmatically. This substitution of measurement methodology fundamentally addresses the spoofing vulnerability while maintaining identification capability.
Solution Approach 2:
The patent introduces sensor data as an intermediary layer between the device and the fingerprinting system. Instead of directly using easily manipulable software identifiers, the system collects physical sensor readings that serve as a more secure intermediary representation of device characteristics, making the fingerprinting process more resistant to manipulation.
2Reliability
If multiple sensor data points are collected and processed through neural networks, then fingerprint robustness is improved, but system complexity increases
Solution Approach 1:
The patent segments the complex fingerprinting process into distinct functional modules: sensor data collection, time-domain feature extraction, frequency-domain feature extraction, and neural network processing. This segmentation allows each component to be optimized independently and simplifies the overall system architecture while maintaining high robustness through the combination of multiple processing stages.
Solution Approach 2:
The patent transforms raw sensor data into multiple feature dimensions by extracting both time-domain and frequency-domain characteristics. This dimensional transformation enriches the data representation without requiring a proportional increase in processing complexity, as the neural network learns to operate efficiently in this transformed feature space.
Data Source
AI summary
Methods and systems are presented for generating a device fingerprint based on data obtained from one or more sensors on a device. A plurality of data points corresponding to sensor readings are obtained from the one or more sensors on the device. A set of time-domain features and a set of frequency-domain features are extracted from the plurality of data points and inputted to a neural network trained using a triplet network. A device fingerprint that may be used to identify the device is obtained from the neural network.