A Comprehensive Survey on Local Differential Privacy toward Data Statistics and Analysis

Area of Science:

• Computer Science
• Information Security
• Data Privacy

Background:

• Crowdsensing utilizes smart device data for decision-making, but raises significant user privacy concerns.
• Existing privacy models struggle to balance data utility with robust individual privacy protection.

Purpose of the Study:

• To provide a comprehensive overview of Local Differential Privacy (LDP) as a privacy-preserving technique in crowdsensing.
• To systematically analyze LDP models, mechanisms, and applications for secure data collection and analysis.

Main Methods:

• Theoretical summarization of LDP models, variants, and algorithmic frameworks.
• Investigation and comparison of LDP mechanisms for frequency estimation, mean estimation, and machine learning tasks.
• Review of practical LDP-based application scenarios.

Main Results:

• LDP offers strong privacy guarantees by perturbing data locally on the client-side.
• Diverse LDP mechanisms are evaluated for their effectiveness in various data analysis tasks.
• Practical applications demonstrate the feasibility and benefits of LDP in real-world scenarios.

Conclusions:

• LDP is a crucial privacy model for crowdsensing, safeguarding user data effectively.
• Future research should focus on advancing LDP mechanisms and exploring new application domains.