Privacy-Preserving Graph Machine Learning from Data to Computation: A Survey

Area of Science:

• Artificial Intelligence
• Data Science
• Machine Learning

Background:

• Data collection, sharing, and analysis in graph machine learning (GML) involve multiple parties with diverse security needs.
• Preserving privacy is crucial for protecting sensitive information within complex, big data networks.
• Graph data structures and graph-based AI models, like graph neural networks, are increasingly used in various domains.

Purpose of the Study:

• To systematically review existing privacy-preserving techniques in graph machine learning.
• To categorize and analyze methods from both data generation and computational aspects.
• To identify current challenges and future research directions in secure GML.

Main Methods:

• Comprehensive literature review of privacy-preserving methods in GML.
• Categorization of techniques based on data generation and information transmission.
• Analysis of theoretical methodologies, software tools, and practical applications.

Main Results:

• Methods for generating privacy-preserving graph data are reviewed.
• Techniques for secure transmission of graph model parameters are described for distributed computation.
• Discussion includes theoretical foundations, software tools, and challenges in the field.

Conclusions:

• The review provides a structured overview of privacy-preserving GML techniques.
• Identified challenges and future research opportunities aim to advance secure GML systems.
• Envisions a unified and comprehensive secure graph machine learning system.