Differential Privacy for the Vast Majority.

Area of Science:

• Computer Science
• Data Privacy
• Database Systems

Background:

• Differential privacy is a standard for protecting sensitive data in databases.
• Existing differential privacy methods assume one record per individual, which is often not true.
• This assumption can lead to significant information loss when individuals have multiple records.

Purpose of the Study:

• To extend differential privacy to databases with multiple records per individual.
• To propose a novel privacy principle integrating differential privacy with the Pareto principle.
• To reduce information loss in query results without compromising privacy.

Main Methods:

• Developed a new privacy principle combining differential privacy and the Pareto principle.
• Applied the extended principle to scenarios with multiple records per person.
• Evaluated the approach using three real-world databases.

Main Results:

• The proposed approach significantly reduces information loss in query results.
• Achieved this reduction with only a small relaxation of the differential privacy guarantee.
• Demonstrated effectiveness across diverse real-world datasets.

Conclusions:

• The integrated privacy principle offers a practical solution for multi-record databases.
• This method enhances data utility while preserving robust privacy.
• It provides a valuable advancement for privacy-preserving data analysis.