Crowd-Sourced Event Identification Maintaining Source Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for crowd-sourced analytics face challenges in ensuring user privacy while handling large volumes of data from mobile devices, as they often rely on encryption techniques that can be vulnerable to data breaches and fail to effectively abstract location data, especially in scenarios involving clusters of behaviors or phenomena across regions.
Innovation Solution
A system that decouples identity information from event reporting using a two-part process, employing on-device analytics to generate notifications and a cloud-based analytics server that utilizes k-medoid clustering and multi-resolution grid-cell clustering to create summary nodes, which are then associated with geographical features, thereby preserving user anonymity and scaling to handle event storms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If encryption techniques are used to protect data-in-transit and data-at-rest, then data privacy is improved, but the system becomes vulnerable to data breaches and cannot effectively abstract location data
Solution Approach 1:
The patent extracts location data from the original event notifications and processes it separately through abstractors that generate abstract location data. This separation removes the vulnerable element (detailed location information) from the main data flow, allowing encryption to protect only the essential identifiers while the abstracted location data provides privacy without relying solely on encryption strength.
Solution Approach 2:
The system segments the authentication and data reporting process into two distinct parts: (1) authentication of device identity, and (2) reporting of event data. This segmentation allows identity verification to occur through secure authentication mechanisms while the actual event data can be transmitted and processed with different privacy protections, reducing the risk that a single breach compromises both identity and data.
2Measurement precision
If polling processes provide constant position and sensor data updates, then event detection accuracy is improved, but user privacy is compromised as data can be assembled into tracks
Solution Approach 1:
The system performs preliminary abstraction of location data on the device side before transmission. Abstractors process raw location data and convert it into abstract location data that retains sufficient information for event detection while removing personally identifiable location patterns. This preliminary action ensures privacy protection is built into the data structure before it leaves the device.
Solution Approach 2:
The patent introduces abstractors as intermediary components between the raw sensor data and the analytics processing system. These abstractors act as mediators that transform detailed location information into abstract representations, allowing the analytics system to perform accurate event detection without accessing or storing detailed personal location tracks.
3Reliability
If random identifiers are used to abstract location data, then privacy is improved, but the system cannot effectively analyze clusters of behaviors or phenomena across regions
Solution Approach 1:
The patent applies different levels of abstraction to different types of location data based on their analytical needs. Some location data is abstracted to protect individual privacy, while aggregated location patterns across multiple devices are preserved to enable regional analysis of behavior clusters. This local quality approach allows the system to maintain privacy for individuals while preserving analytical capabilities for population-level insights.
Data Source
AI summary
An automated method that generates crowd-sourced event notifications includes: retrieving sensor data from mobile devices; detecting events based on the retrieved sensor data; and generating a set of summary nodes based on the events, where each summary node is associated with multiple events. A system that generates crowd-sourced event notifications includes: multiple mobile devices, each having at least one sensor, each mobile device identifies events using calculations based on sensor data and generates a notification for each identified event; and an analytics server that generates a set of summary nodes based on the identified events, each summary node associated with multiple events. An automated method that generates event notifications includes: retrieving sensor data; calculating a set of statistical values based on the sensor data; comparing the statistical values to at least one threshold; and generating an event notification when the threshold is exceeded by any of the statistical values.