Database Reference
In-Depth Information
Table 9.2: Frequent itemsets for database D
O
at mfreq = 0:3.
Frequent itemsets
Support
{a}
7
{b}, {c}
6
{ac}
5
{d}, {e}, {ab}, {bc}
4
{ad}, {ae}, {be}, {cd}, {ce}, {abc}, {acd}, {ace}
3
Fig. 9.1: An itemsets' lattice demonstrating (i) the original border and the sensitive
itemsets, and (ii) the revised border.
sets, assume that S = fe; ae; bcg are the sensitive ones that have to be protected.
Figure 9.1
demonstrates the process of border revision for the problem at hand. In
this figure, near each itemset we depict its support in the original
(Figure 9.1(i))
and
the sanitized (
Figure 9.1(ii))
database, which facilitates optimal hiding of the sensi-
tive knowledge. Notice that although the sanitized database is not yet constructed,
the process of border revision allows us to compute the borderline of the optimal
hiding solution. As one can observe, there are four possible scenarios involving the
status of an itemset I prior and after the application of border revision:
C
1
Itemset I was frequent in D
O
and remains frequent in D.
C
2
Itemset I was infrequent in D
O
and remains infrequent in D.
C
3
Itemset I was frequent in D
O
and became infrequent in D.
