Trained User Behavior Model for Secure Service Personalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data security measures in machine learning and user data analysis expose user actions to unauthorized access by transmitting personal data for identification, compromising information security.
Innovation Solution
Generating individual content for services using a trained model of user behavior, rather than transmitting actual personal data, to ensure secure user identification and data protection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If personal user data is transmitted for user identification and analysis, then user identification accuracy is improved, but information security deteriorates due to exposure to unauthorized access
Solution Approach 1:
The patent extracts only the essential identification characteristics from personal data and transmits them to remote devices for analysis, while keeping the complete personal data set local on user devices. This extraction approach enables user identification without exposing sensitive personal information to external systems, thereby resolving the contradiction between identification accuracy and information security.
Solution Approach 2:
The patent introduces trained machine learning models as intermediaries that process personal data locally on user devices and generate condensed identification representations. These models act as mediators between personal data and remote analysis systems, enabling accurate user identification while preventing direct exposure of personal data to external systems.
2Object-affected harmful factors
If personal data is stored locally on user devices to protect information security, then information security is improved, but the ability to provide personalized services deteriorates due to limited data access
Solution Approach 1:
The patent applies preliminary action by training machine learning models locally on user devices using stored personal data before any data transmission occurs. These pre-trained models enable personalized service delivery through local inference and generate condensed user representations for remote interactions, allowing personalized services to be provided without exposing personal data to external systems.
Solution Approach 2:
The patent creates local copies of user behavior patterns and preferences through trained machine learning models stored on user devices. These model copies enable personalized service generation locally without requiring access to actual personal data, thereby maintaining information security while preserving personalized service capability.
3Object-affected harmful factors
If trained models are used to process personal data locally, then information security is improved by preventing data transmission, but device resource consumption increases due to model training and execution
Solution Approach 1:
The patent applies partial action by implementing model training and processing only on devices with sufficient computational resources. User devices that cannot support full model training perform lighter processing or rely on pre-trained models, while more powerful devices handle the computationally intensive training tasks. This selective approach enables local data processing for information security without universally increasing resource consumption across all devices.
Data Source
AI summary
Disclosed herein are systems and methods for generating individual content for a user of a service. In one aspect, an exemplary method comprises, gathering data on behavior of a user of a computing device, training a model of a user behavior based of the gathered data, wherein the trained data identifies the user to a predetermined degree of reliability, and generating an individual content for the user of the service based on a predetermined service environment in accordance with a trained model received from a model transmitter.