User Identification via Activity Matrix Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for identifying users in a network, such as using IP addresses and tracking cookies, are unreliable and raise privacy concerns, as they fail to accurately distinguish between users accessing from different devices or locations, and may infringe on user privacy.
Innovation Solution
A server processes user activity data to create a user activity matrix, applies a transformation model to predict new user activity, and compares the results to identify unique users based on behavior patterns without using personally identifiable information, employing techniques like random forests and compact singular value decomposition for dimensionality reduction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If IP addresses and tracking cookies are used for user identification, then the access granting process is simplified, but the reliability of user identification deteriorates due to inability to distinguish users accessing from different devices or locations
Solution Approach 1:
The patent transforms the user identification approach by changing from direct identification parameters (IP addresses, cookies) to behavioral pattern parameters. It applies matrix transformation operations that convert raw activity data into transformed representations capturing essential user behavior characteristics, thereby improving identification reliability while maintaining operational simplicity
Solution Approach 2:
The patent introduces a matrix transformation model as an intermediary between raw activity data and user identification. This model acts as a mediator that processes activity matrices through mathematical transformations (including dimensionality reduction techniques) to produce reliable user identification without directly relying on traditional tracking mechanisms
2Ease of operation
If IP addresses and tracking cookies are used for user identification, then access control is simplified, but user privacy is compromised due to collection of personally identifiable information
Solution Approach 1:
The patent extracts only the essential behavioral patterns from user activity data while discarding personally identifiable information. By focusing on transforming activity matrices to capture behavioral essences rather than storing or processing raw PII, it achieves privacy-preserving user identification that maintains access control functionality
Solution Approach 2:
The patent changes the identification parameters from PII-based (IP addresses, cookies) to behavioral pattern-based representations. The matrix transformation operations convert activity data into abstracted forms that preserve identification capability while eliminating direct links to user identity, thereby protecting privacy
3Device complexity
If traditional user identification methods are used, then implementation is simple, but accuracy in distinguishing unique users deteriorates
Solution Approach 1:
The patent segments the user identification problem into distinct processing stages: collecting activity data, forming activity matrices, applying transformation operations, and performing identification. This segmentation allows the system to manage complexity through structured processing steps while achieving high identification accuracy through specialized operations at each stage
Solution Approach 2:
The patent introduces matrix transformation operations as intermediary processing steps between data collection and identification. These transformations (including dimensionality reduction) serve as mediating operations that enhance measurement precision by extracting meaningful patterns while managing computational complexity through structured mathematical operations
Data Source
AI summary
Techniques of identifying users involve automatically determining whether new user activity data received by a server is associated with a user known to the server. Along these lines, a server collects data representing network activity of a group of users. The data collected takes the form of a table, or matrix, with each entry corresponding to a respective user and having values of a respective set of fields. The server may then use this data as training data in developing a model for predicting whether new activity data corresponds to one of the group of users or a new user not in the group of users.