Masking User Input and Sensor Data to Prevent Behavioral Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for tracking user behavior at a user device, such as typing tendencies and sensor data, can still identify users even when they have enabled privacy modes or disabled data monitoring services, as these methods do not effectively mask the unique patterns in user input and sensor readings.
Innovation Solution
Implementing a system that normalizes and randomizes user input and sensor data to prevent identification, by intercepting and modifying the data to ensure consistent and randomized delays or readings, thereby making it difficult for services to uniquely identify users based on their behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If privacy modes or data monitoring services are disabled, then user activity tracking is reduced, but unique identification through typing tendencies and sensor data patterns remains possible
Solution Approach 1:
The system changes the parameters of user input data by introducing random delays between key presses and random variations in typing patterns. This transforms the original typing pattern parameters (timing, rhythm) into modified parameters that preserve functionality while eliminating unique identification characteristics. The sensor data parameters (orientation, acceleration) are similarly randomized to prevent pattern recognition.
2Productivity
If user input data is collected for service functionality, then service performance is maintained, but user identification through pattern analysis becomes possible
Solution Approach 1:
The system introduces an intermediary masking layer between the user input and the service processing. This intermediary component randomizes and modifies the input data before it reaches the service, allowing the service to function with modified data while preventing direct access to original user patterns. The intermediary ensures that services receive usable data without exposing identifying characteristics.
3Ease of operation
If sensor data is collected for device functionality, then device operation is maintained, but unique user identification through sensor patterns remains possible
Solution Approach 1:
Instead of collecting original sensor data and attempting to anonymize it later, the system inverts the approach by immediately transforming sensor data into randomized patterns upon collection. The inversion occurs at the data generation stage rather than the processing stage, ensuring that original patterns never persist in the system. Sensor readings are replaced with randomized equivalents that maintain functional utility but eliminate identifying patterns.
Data Source
AI summary
A system described herein may allow for the masking of user input and/or sensor data, which could otherwise be used to uniquely identify and track a user. For example, user inputs (e.g., keyboard or mouse inputs) and/or sensor data (e.g., data from a touchscreen, pressure sensor, gyroscope, etc.) may be normalized and randomized. The normalization and/or randomization may include modifying metadata associated with user inputs or sensor data (e.g., modification of timestamps and/or modification of raw data) prior to outputting the user inputs or sensor data to an application, and/or to a service that attempts to uniquely identify users based on such metadata.