Interlayer Link Prediction in Multiplex Social Networks Based on Multiple Types of Consistency Between Embedding...

Area of Science:

• Computer Science
• Network Science
• Cybersecurity

Background:

• Users active on multiple social media networks (SMNs) create complex multiplex social networks.
• Increasing cybersecurity awareness leads users to adopt distinct profiles across SMNs, complicating user identification.
• Linking accounts across different SMNs is a challenging interlayer link prediction problem.

Purpose of the Study:

• To propose a novel framework, Multiple types of Consistency between Embedding Vectors (MulCEV), for accurate interlayer link prediction in multiplex social networks.
• To address limitations of existing embedding-based methods by incorporating positional relationships and weighted graph information.
• To improve the accuracy of identifying the same user across different social media platforms.

Main Methods:

• Leveraging network embedding techniques to learn latent representations of nodes within each social media network layer.
• Introducing a distance consistency index based on node positions in latent spaces to enhance prediction clues.
• Modeling social media layers as weighted graphs to capture relationship strengths, unlike unweighted approaches.
• Combining traditional embedding consistency with the proposed distance consistency for a comprehensive prediction approach.

Main Results:

• The MulCEV framework significantly outperforms existing embedding-based methods in interlayer link prediction.
• MulCEV demonstrates superior performance, particularly with a limited number of training iterations.
• Experiments on real-world and synthetic datasets validate the effectiveness of the proposed approach.

Conclusions:

• The MulCEV framework offers a robust solution for the challenging problem of user identification across multiplex social networks.
• By integrating multiple consistency measures and weighted graph modeling, MulCEV fully utilizes latent space information.
• This approach enhances cybersecurity by improving the ability to accurately link user accounts across diverse online platforms.