Training Label Generator for Cross-Device Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for accurately identifying distinct physical user devices in a cross-device context are hindered by duplication of device records due to factors like cookie expiration and multiple browser usage, leading to inaccurate analytics and the need for expensive and scarce labeled training data for supervised learning.
Innovation Solution
A multi-phase process for generating labeled training data using a training label generator, which filters out trivial device record pairs, generates positive and negative labels based on shared authentication events and temporal activity data, and leverages external data sources to create robust training datasets for classifiers to differentiate between equivalent and non-equivalent device records.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning is used to train classifiers to identify distinct physical user devices, then prediction accuracy can be improved, but the cost and time required to generate labeled training data increases significantly
Solution Approach 1:
The system performs preliminary actions by collecting and storing device attributes, authentication events, and activity data in advance during normal operation. This pre-collected data is then used to automatically generate labeled training pairs without requiring time-consuming manual annotation, thus improving prediction accuracy while reducing data preparation time
Solution Approach 2:
The system enables self-service by automatically generating labeled training data through unsupervised learning mechanisms. The classifier trains itself using device records and their inherent characteristics (cookies, hardware identifiers, authentication patterns) without human intervention, eliminating the need for expensive manual labeling while maintaining high accuracy
2Reliability
If more labeled training data is collected to improve classifier robustness, then prediction reliability increases, but the complexity of data management and processing increases
Solution Approach 1:
The system segments the training data generation process into distinct components: device attribute collection, authentication event tracking, activity data recording, and automatic label generation. This segmentation allows each component to be managed independently, reducing overall data management complexity while enabling comprehensive data collection for improved classifier reliability
Solution Approach 2:
The system creates a multi-functional data collection framework that simultaneously serves multiple purposes: training data generation, device fingerprinting, authentication verification, and analytics. This universal approach consolidates data management operations into a single system, reducing complexity while providing robust training data for reliable predictions
3Measurement precision
If device records are differentiated based on multiple attributes to improve identification accuracy, then the ability to distinguish distinct devices improves, but the computational resources required for processing increases
Solution Approach 1:
The system applies partial action by selectively using the most discriminative device attributes for training and prediction, rather than processing all possible attributes equally. This approach achieves high device differentiation precision while reducing computational energy consumption by focusing only on the most relevant features
Solution Approach 2:
The system changes parameters by transforming raw device attributes into normalized feature representations and adjusting the weighting of different attributes based on their discriminative power. This parameter optimization enables accurate device differentiation while minimizing computational resource requirements through efficient feature engineering
Data Source
AI summary
Techniques are disclosed for accurately identifying distinct physical user devices in a cross-device context. An example embodiment applies a multi-phase approach to generate labeled training datasets from a corpus of unlabeled device records. Such labeled training datasets can be used for training machine learning systems to predict the occurrence of device records that have been wrongly (or correctly, as the case may be) attributed to different physical user devices. Such identification of improper attribution can be particularly helpful in web-based analytics. The labeled training datasets include labeled pairs of device records generated using multiple strategies for inferring whether the two device records of a pair of device records represent the same physical user device (or different physical user devices). The labeled pairs of device records can then be used to train classifiers to predict with confidence whether two device records represent or do not represent the same physical user device.