A clustering-based differential privacy protection algorithm for weighted social networks.

Area of Science:

• Computer Science
• Data Privacy
• Network Analysis

Background:

• Weighted social networks are vital in diverse applications like social media and healthcare.
• Increasing use raises significant privacy concerns, including sensitive data leakage and privacy attacks.
• Existing differential privacy methods struggle with data utility due to excessive noise in edge weights.

Purpose of the Study:

• To propose a novel privacy protection algorithm, DCDP, for weighted social networks.
• To address the trade-off between privacy preservation and data utility in weighted network analysis.
• To safeguard sensitive user information while maintaining data accuracy.

Main Methods:

• Developed the DCDP algorithm combining OPTICS density clustering with differential privacy.
• Partitioned weighted networks into sub-clusters for targeted noise injection.
• Introduced a new privacy parameter calculation method for balanced protection.

Main Results:

• DCDP achieves differential privacy for weighted social networks while preserving data accuracy.
• Reduced average relative error by approximately 20% compared to traditional methods.
• Increased the proportion of unchanged shortest paths by about 10%.

Conclusions:

• DCDP offers an effective solution for privacy protection in weighted social networks.
• The algorithm successfully balances robust privacy with high data utility.
• DCDP provides a valuable tool for secure analysis of sensitive network data.