Database Reference
In-Depth Information
Chapter 5
Summary
Association rule hiding is a subarea of privacy preserving data mining that focuses
on the privacy implications originating from the application of association rule min-
ing to large public databases. In this first part of the topic, we provided the basics
for the understanding of the problem, which investigates how sensitive association
rules can escape the scrutiny of malevolent data miners by modifying certain val-
ues in the original database, and presented some related problems on the knowledge
hiding thread. Specifically, in the first two chapters we motivated the problem of as-
sociation rule hiding, presented the necessary background for its understanding and
derived the problem statement along two popular variants of the problem: frequent
itemset hiding and association rule hiding. In Chapter 3, we provided a classification
of the association rule hiding algorithms to facilitate the organization that we follow
for the presentation of the methodologies in the rest of this topic. Our proposed tax-
onomy partitions the association rule hiding methodologies along four orthogonal
directions based on the employed hiding strategy, the data modification strategy, the
number of rules that are concurrently hidden, and the nature of the algorithm. Elab-
orating on the last direction, we identified three classes of association rule hiding
approaches, namely heuristic-based, border-based and exact approaches, and dis-
cussed the differences among them. Last, in Chapter 4 we examined the problem
of knowledge hiding in the related research areas of clustering, classification and
sequence mining. For each of these areas we briefly discussed some state-of-the-art
approaches that have been proposed.
