Semantic attack on anonymised transactions

Abstract

A transaction is a data record that contains items associated with an individual. For example, a set of movies rated by an individual form a transaction. Transaction data are important to applications such as marketing analysis and medical studies, but they may contain sensitive information about individuals which must be sanitised before being used. One popular approach to anonymising transaction data is set-based generalisation, which attempts to hide an original item by replacing it with a set of items. In this paper, we study how well this method can protect transaction data. We propose an attack that aims to reconstruct original transaction data from its set-generalised version by analysing semantic relationships that exist among the items. Our experiments show that set-based generalisation may not provide adequate protection for transaction data, and about 50 % of the items added to the transactions during generalisation can be detected by our method with a precision greater than 80%.