Adaptive LDP Framework for Numerical Data Collection Without Prior Domain Knowledge
A new study on arXiv (2608.05737) presents a fresh approach to Local Differential Privacy (LDP) that addresses the challenge of collecting numerical data without prior knowledge of the data category. This method requires users to send two pieces of information: their modified numerical data and a privatized signal indicating whether their original value was clipped. By gathering these signals, the framework can adjust the domain range, which helps reduce information loss from clipping and limits excessive noise from wide domains. Consequently, this improves the quality of the data collected while maintaining strong privacy safeguards. The paper is categorized as a cross-announcement on arXiv.
Key facts
- The framework is designed for Local Differential Privacy (LDP).
- It addresses the challenge of unknown data domains in numerical data collection.
- Each user sends perturbed numerical data and a privatized clipping signal.
- The method dynamically estimates the data domain.
- It reduces information loss from clipping and excessive noise.
- The paper is available on arXiv with ID 2608.05737.
- The announcement type is cross.
- The paper was announced on arXiv.
Entities
Institutions
- arXiv