Set-valued data collection with local differential privacy based on category hierarchy.

Area of Science:

• Computer Science
• Information Security

Background:

• Set-valued data is crucial in sensor technology.
• Existing local differential privacy (LDP) methods for set-valued data suffer from information loss and ignore category privacy.
• The trustworthiness assumption of centralized differential privacy (DP) necessitates LDP solutions.

Purpose of the Study:

• To propose a novel set-valued data collection method (SetLDP) under the local differential privacy (LDP) model.
• To address the information loss and category privacy issues in existing LDP methods for set-valued data.
• To enhance data utility while ensuring robust privacy protection for set-valued data categories.

Main Methods:

• Developed SetLDP, a method based on category hierarchy for LDP set-valued data collection.
• Implemented a privacy mechanism involving random response to category existence, perturbation of item counts, and a new utility function for candidate itemsets.
• Utilized theoretical analysis and experimental validation to evaluate the proposed method.

Main Results:

• SetLDP effectively preserves more data utility compared to existing methods.
• The proposed method successfully protects the private information of set-valued data categories.
• Experimental results validate the efficacy of SetLDP in balancing privacy and utility.

Conclusions:

• SetLDP offers an improved approach for privacy-preserving set-valued data collection under LDP.
• The method enhances data utility and provides robust protection for category-level privacy.
• SetLDP is a promising solution for sensitive set-valued data applications in sensor networks.