Application of hyperbolic geometry of multiplex networks under layer link-based attacks

Area of Science:

• Network Science
• Complex Systems
• Data Analysis

Background:

• Multiplex networks, comprising interconnected layers, are crucial for modeling complex systems.
• Identifying interactions and predicting link behavior in these networks is an active research area.
• Existing link prediction methods face challenges in accurately assessing redundancy in multilayered structures.

Purpose of the Study:

• To propose a novel method for predicting redundant links in multiplex networks.
• To enhance the resilience of multiplex networks against link loss and attacks.
• To leverage hyperbolic geometry for improved link prediction accuracy.

Main Methods:

• A new link prediction method based on hyperbolic distance similarity criterion for node pairs.
• Application of the method to identify and retrieve lost links in multiplex networks.
• Evaluation against existing algorithms using real-world network data and simulations.

Main Results:

• The proposed hyperbolic distance method demonstrates superior performance in predicting redundant links.
• Effective retrieval of lost links under diverse attack strategies, including edge betweenness and random attacks.
• Numerical simulations confirm the method's advantage over standard and advanced link prediction techniques.

Conclusions:

• Hyperbolic geometry provides a powerful framework for link prediction in multiplex networks.
• The developed method offers enhanced network retrieval capabilities, particularly against targeted attacks.
• This approach advances the understanding and robustness of multilayered network analysis.