Confidentiality of the Patient Records and the Data Inference Problem
Ljiljana Branković, Maddison Miller, L Bomba, Joanne E. Cooper, Stuart Morris, Fernando Fonseca Pacheco · RUNE (Research UNE) · 1995
The problem of data inference from statistical queries (such as MAX, MIN.AVG, SUM and COUNT) is particularly pertinent to medical databases. A medical database is used by individual doctors each of whom has an unrestricted access to their own patients' records. Additionally, medical researchers and health authorities have access to the answers to statistical queries about subpopulations of the database but not to any of the individual records. However, it is often possible to infer information about an individual from the answers to statistical queries by correlating the obtained statistics (possibly supplemented by other information obtained from within or without the database). If this happens, we say that the database has been compromised. At present, none of the commercially available database management systems offer any form of protection from data inference due to statistical queries. The situation is likely to change as more medical information systems become computerised and integrated. In this paper we present an overview of the data inference problem, the most common data inference techniques and the available mechanisms for the prevention of a database compromise