Disclosure Analysis and Control in Statistical Databases

  • Authors:
  • Yingjiu Li;Haibing Lu

  • Affiliations:
  • Singapore Management University, Singapore 178902;Rutgers University, Newark, NJ 07102

  • Venue:
  • ESORICS '08 Proceedings of the 13th European Symposium on Research in Computer Security: Computer Security
  • Year:
  • 2008

Quantified Score

Hi-index 0.00

Visualization

Abstract

Disclosure analysis and control are critical to protect sensitive information in statistical databases when some statistical moments are released. A generic question in disclosure analysis is whether a data snooper can deduce any sensitive information from available statistical moments. To address this question, we consider various types of possible disclosure based on the exact bounds that a snooper can infer about any protected moments from available statistical moments. We focus on protecting static moments in two-dimensional tables and obtain the following results. For each type of disclosure, we reveal the distribution patterns of protected moments that are subject to disclosure. Based on the disclosure patterns, we design efficient algorithms to discover all protected moments that are subject to disclosure. Also based on the disclosure patterns, we propose efficient algorithms to eliminate all possible disclosures by combining a minimum number of available moments. We also discuss the difficulties of executing disclosure analysis and control in high-dimensional tables.