Privacy-Preserving Knowledge Graph Extension via AI Intermediaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a demand for AI technology that effectively protects a user's privacy while extending a knowledge graph related to the user.
Innovation Solution
A system and method that utilize multiple artificial intelligence (AI) models to obtain, extend, and use a knowledge graph related to a user, while protecting individual privacy by abstracting data according to a set privacy level and integrating context information from both device and server sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a knowledge graph is extended by collecting more user data, then the accuracy and personalization of recommended content is improved, but user privacy protection deteriorates
Solution Approach 1:
The patent introduces a privacy-preserving intermediary layer that mediates between user data and the knowledge graph extension process. This intermediary anonymizes and abstracts user information before it is used to build the knowledge graph, allowing the system to extend the knowledge graph with sufficient detail for accurate recommendations while preventing direct exposure of sensitive user data. The intermediary transforms personal information into aggregated patterns that maintain utility for recommendation accuracy without compromising individual privacy.
2Reliability
If multiple AI models are used to extend the knowledge graph, then the quality and comprehensiveness of the knowledge graph is improved, but the system complexity increases
Solution Approach 1:
The patent divides the knowledge graph extension task into multiple specialized AI models, each responsible for specific functions such as entity recognition, relation extraction, and knowledge fusion. This segmentation allows each model to be optimized for its specific function, improving overall knowledge graph quality while making the complex system more manageable through modular architecture. Each segmented model processes specific aspects of data, reducing the complexity burden on any single component.
Solution Approach 2:
The patent introduces intermediary modules that coordinate between multiple AI models, managing data flow and integration. These intermediaries handle the complexity of orchestrating multiple models by providing standardized interfaces and coordination mechanisms, allowing the system to leverage multiple AI models for high-quality knowledge graph extension while abstracting away the operational complexity from the overall system management.
3Loss of information
If context information is collected from both device and server sources, then the completeness of the knowledge graph is improved, but the risk of privacy leakage increases
Solution Approach 1:
The patent applies different privacy protection strategies to different sources of context information based on their sensitivity and nature. Device-collected information receives one level of protection while server-provided information receives another level, allowing the system to maximize knowledge graph completeness by selectively integrating information from both sources while maintaining appropriate privacy safeguards for each source according to its specific characteristics.
Solution Approach 2:
The patent introduces privacy-aware intermediary layers that selectively filter and process context information from device and server sources. These intermediaries evaluate each information source against privacy criteria and only allow non-sensitive or appropriately anonymized data to contribute to the knowledge graph, thereby maintaining completeness where possible while blocking privacy-leaking information paths.
Data Source
AI summary
Provided are an artificial intelligence (AI) system using a machine learning algorithm and an application of the AI system. A device for providing content based on a knowledge graph includes: a memory storing instructions; and a processor configured to execute the instructions to: obtain context information related to the device; obtain a first device knowledge graph of a user of the device by inputting the obtained context information to a first AI model for determining a relation between entities related to the user of the device; request, from a server, a server knowledge graph generated by the server; receive the server knowledge graph; obtain a second device knowledge graph of the user by inputting the obtained first device knowledge graph and the received server knowledge graph to a second AI model for extending the first device knowledge graph; and provide content based on the obtained second device knowledge graph.