The number of unique records in a released microdata set which are unique in the population is an important measure of re-identification disclosure risk in microdata. However, the microdata sample contains information about the disclosure risk more than the number of unique records. This paper deals with the development of a technique based on a loglinear models to extract more information from the sample about the disclosure risk not only through the number of sample unique records but also through the number of twin and triple records. These information may help microdata release committee in taking decision about releasing the data for public use. For illustration we apply the proposed method to data from a General Household Survey 1996-1997. © Springer-Verlag 2004.
Elamir, E. A. H. (2004). Analysis of re-identification risk based on log-linear models. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3050, 273–281. https://doi.org/10.1007/978-3-540-25955-8_21