Remote Graph Statistics for Privacy-Safe ML Graph Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graph machine learning models lack accuracy and robustness due to data privacy concerns and varying enforcement of data security policies, leading to inadequate data sharing and insufficient data augmentation.
Innovation Solution
A system and method for remote statistical generation of graphs, where a first electronic device captures anonymized metadata from a graph database, determines statistical information, and transmits it to a second device for generating a representative graph, ensuring data privacy and accuracy by validating the generated graph at the local site.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data masking is used to create an alternate version of data, then data privacy is protected, but the amount of information transfer is not reduced and data augmentation is not helped
Solution Approach 1:
The patent extracts only the essential statistical properties (node counts, edge counts, degree distributions, clustering coefficients) from the original graph data, separating the privacy-sensitive detailed information from the useful structural characteristics needed for machine learning training.
Solution Approach 2:
The patent transforms the original graph data by changing its representation from detailed node-edge structures to aggregated statistical parameters, thereby preserving the essential topological information while eliminating privacy risks and reducing data volume.
2Measurement precision
If graph data is shared outside the service provider for model generation, then machine learning model accuracy can be improved, but data security policies enforcement laxity varies and privacy issues arise
Solution Approach 1:
The patent introduces statistical parameters as an intermediary representation that bridges the gap between data privacy requirements and machine learning model training needs, allowing external partners to work with anonymized statistical data rather than sensitive original data.
Solution Approach 2:
The patent creates a statistical copy of the graph data that replicates the essential structural properties without containing any sensitive information, enabling external partners to train models on this anonymized copy.
3Adaptability or versatility
If detailed graph data is transmitted to external partners, then data augmentation is possible, but the risk of reverse engineering and privacy breaches increases
Solution Approach 1:
The patent extracts only the essential statistical properties (node counts, edge counts, degree distributions, clustering coefficients) from the original graph data, separating the privacy-sensitive detailed information from the useful structural characteristics needed for machine learning training.
Data Source
AI summary
According to an aspect of an embodiment, operations may include retrieving a first graph. The operations may further include identifying a set of node-types, determining a first count of each of the identified set of node-types, and determining first statistical information. The operations may further include identifying a set of edge-types, determining a second count of each of the identified set of edge-types and determining a two-dimensional (2D) distribution of each of the identified set of edge-types. The operations may further include determining second statistical information, identifying a set of combinations of edge-types connecting three node-types and determining a third count of each of a set of three node-type groups. The operations may further include determining a three-dimensional (3D) distribution of each of the set of three node-type groups, determining third statistical information, and transmitting first graph statistics associated with the retrieved first graph for generation a second graph.