Database Reference
In-Depth Information
maximize (u 51 +u 52 +u 81 +u 82 )
<
:
u 52 +u 82 0:6
u 51 0:6
Y 1 u 51
Y 1 u 52
Y 2 u 81
Y 2 u 82
Y 1 u 51 +u 52 1
Y 2 u 81 +u 82 1
Y 1 +Y 2 < 1:6
where fu 51 ; u 52 ; u 81 ; u 82 ;Y 1 ;Y 2 g2f0; 1g
subject to
Fig. 14.4: The CSP formulation for dataset D O .
Table 14.4: The characteristics of the three datasets.
Dataset
N
M
Avg tlen
BMS-WebView-1
59,602
497
2.50
BMS-WebView-2
77,512
3,340
5.00
Mushroom
8,124
119
23.00
two new temporary binary variables, Y 1 and Y 2 , as follows:
8
<
Y 1 u 51 ; Y 2 u 81
Y 1 u 52 ; Y 2 u 82
Y 1 u 51 +u 52 1; Y 2 u 81 +u 82 1
Y 1
z }| {
u 51 u 52 +
Y 2
z }| {
u 81 u 82 < 1:6 )
:
By using this process, all constraints of the CSP that involve products of binary vari-
ables become linear. The resulting CSP is presented i n Figure 14.4 and is solved by
using BIP. The solution of the CSP leads to three optimal hiding solutions, presented
in Table 14.3, among which one is selected to derive the sanitized database D.
14.3 Experiments and Results
The inline algorithm has been tested on real world datasets using different param-
eters such as minimum support threshold and number/size of sensitive itemsets to
hide [23]. In this section, we present the datasets that were used, their special char-
acteristics, the selected parameters and the attained experimental results.
 
Search WWH ::




Custom Search