Database Reference
In-Depth Information
Chapter 12
Summary
In the third part of the topic, we examined two popular border-based approaches
that have been recently proposed for the hiding of sensitive frequent itemsets. The
first approach, called BBA, was presented in Chapter 10 and was developed by Sun
& Yu in [66, 67]. This approach uses the border of the nonsensitive frequent item-
sets to track the impact of altering items of transactions in the original database.
The second approach, called Max-Min, was proposed by Moustakides & Verykios
in [50] and was covered in Chapter 11. It involves two heuristic algorithms that rely
on the max-min criterion and use the revised positive border to hide the sensitive
knowledge with minimal impact on the nonsensitive frequent itemsets. By restrict-
ing the impact of any tentative item modifications to the itemsets of the revised
positive border, both BBA and Max-Min achieve to identify hiding solutions with
fewer side-effects, when compared to pure heuristic approaches. This is due to the
fact that tracking the impact of item modifications to the itemsets of the borders
provides a measure of the side-effects that are induced by the hiding algorithm to
the original database (in terms of lost itemsets) and thus serves as an optimization
criterion to guide the hiding process towards high quality solutions. As a conclud-
ing remark, we should point out that although border-based approaches provide an
improvement over pure heuristic approaches, they are still reliant on heuristics to
decide upon the item modifications that they apply on the original database. As a
result, in many cases these methodologies are unable to identify optimal hiding so-
lutions, although such solutions may exist for the problem at hand.
