Statistical Database Auditing Without Query Denial Threat

• Main Authors: LU, Haibing, VAIDYA, Jaideep, ATLURI, Vijay, LI, Yingjiu
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2014
• Subjects: statistical database; privacy; auditing; query denial; optimization; Computer Sciences; Numerical Analysis and Scientific Computing
• Online Access: https://ink.library.smu.edu.sg/sis_research/2550; http://dx.doi.org/10.1287/ijoc.2014.0607
• Institution: Singapore Management University

Statistical database auditing is the process of checking aggregate queries that are submitted in a continuous manner, to prevent inference disclosure. Compared to other data protection mechanisms, auditing has the features of flexibility and maximum information. Auditing is typically accomplished by examining responses to past queries to determine whether a new query can be answered. It has been recognized that query denials release information and can cause data disclosure. This paper proposes an auditing mechanism that is free of query denial threat and applicable to mixed types of aggregate queries, including sum, max, min, deviation, etc. The core ideas are (i) deriving the complete information leakage from each query denial and (ii) carrying the complete leaked information derived from past answered and denied queries to audit each new query. The information leakage deriving problem can be formulated as a set of parametric optimization programs, and the whole auditing process can be modeled as a series of convex optimization problems.