Knowledge Graph Anonymization Using K-Anonymity and Differential Privacy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for utilizing knowledge graphs waste human and computing resources by violating data privacy legal compliance requirements during data utilization, as there are no frameworks for applying data privacy in the context of knowledge graphs.
Innovation Solution
An anonymization system applies a k-anonymity model to protect node level privacy and a differential privacy model to protect edge level privacy in knowledge graphs, leveraging knowledge graph embeddings to maximize data utility while ensuring compliance with data privacy legal requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If knowledge graphs are utilized without privacy protection, then data utility and computational efficiency are improved, but privacy compliance is violated
Solution Approach 1:
The patent segments the knowledge graph into multiple equivalence classes based on quasi-identifier attributes (e.g., age, gender, occupation). Nodes are divided into classes such that each class contains entities with similar attribute values, ensuring that no single entity can be uniquely identified. This segmentation enables privacy protection while maintaining the structural integrity and utility of the knowledge graph for analytical purposes.
Solution Approach 2:
The patent introduces an intermediary anonymization layer between the original knowledge graph and the output data. This intermediary process applies k-anonymity by creating equivalence classes and merging nodes, while a differential privacy mechanism adds noise to query results. This intermediary layer mediates between data utility requirements and privacy protection, allowing useful insights to be extracted without exposing individual entity information.
2Object-affected harmful factors
If k-anonymity is applied to protect node level privacy, then privacy protection is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-computing equivalence classes based on quasi-identifier attributes before processing queries. The knowledge graph is pre-analyzed to identify nodes that should be merged into the same equivalence class, and these classes are pre-established. This preliminary classification reduces the computational burden during query processing, as the anonymization structure is already in place and can be efficiently applied to new queries without re-computing the entire equivalence classification.
3Object-affected harmful factors
If differential privacy is applied to protect edge level privacy, then privacy protection is improved, but data utility decreases
Solution Approach 1:
The patent applies parameter changes by adjusting the noise addition mechanism in the differential privacy framework. Instead of using a fixed noise level, the system dynamically adjusts privacy parameters (such as the epsilon parameter in (ε, δ)-differential privacy) based on the specific query and data characteristics. This allows the system to optimize the balance between privacy protection and data utility, adding minimal necessary noise to protect edge privacy while preserving meaningful relationships and patterns in the knowledge graph.
Data Source
AI summary
A device processes a knowledge graph to select a set of triples and generates a first class based on the set of triples. The device determines a quantity of quasi-identifier attributes in the first class and compares the quantity to a predefined parameter. The device embeds the knowledge graph to generate an embedding space representation, identifies a second class, and determines a first quantity of nodes in the first class and the second class. The device compares the first quantity to the predefined parameter and identifies a third class. The device determines a second quantity of nodes in the first class and the third class and compares the second quantity to the predefined parameter. The device merges the second class or the third class with the first class, based on the comparisons, to generate anonymized nodes for the knowledge graph.