Optimizing Differential Privacy Guarantees for High-Dimensional Data Publishing

Authors

  • Elijah Thornton , PhD Candidate, Department of Computer Science, Clark Atlanta University, Atlanta, Georgia, United States
  • Naomi Whitfield , PhD Candidate, School of Data Science, University of Arkansas at Little Rock, Arkansas, United States
  • Marcus Delgado , PhD Candidate, Department of Mathematics, Howard University, Washington, United States
  • Linh Nguyen PhD Candidate, Department of Computer Science, New Mexico State University, Las Cruces, New Mexico

Keywords:

Differential Privacy, High-dimensional Data, Utility-Privacy Trade-off, Adaptive Sensitivity Control, Data Publishing

Abstract

Applying differential privacy (DP) to high-dimensional data publishing presents significant challenges in balancing privacy guarantees with data utility. This work introduces a method for optimizing DP in such settings, focusing on practical trade-offs rather than theoretical ideals. The proposed approach combines adaptive sensitivity control with dimensionality-aware noise allocation, aiming to preserve utility for common analytical tasks. Initial results suggest improvements over baseline methods, though limitations persist in scenarios with extreme dimensionality or very strict privacy budgets. Empirical tests on standardized datasets indicate that the method may offer reasonable utility retention for typical use cases, but further validation is required for broader applicability. The computational requirements appear manageable for moderately sized datasets, though scalability to extremely high dimensions remains an open question. This work contributes to ongoing efforts to make DP more practical for real-world data publishing scenarios.

Downloads

Published

2025-04-17

Issue

Vol. 1 No. 2 (2024)

Section

Articles

License

Copyright (c) 2025 Future - Artificial Intelligence and Social Systems: Innovations and Impacts (AISSII)

Creative Commons License

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.