Databases Reference
In-Depth Information
both support and confidence of the appropriate rules could be lowered. In this
case, 0-values in the transactional database could also change to 1-values. In
many cases, this resulted in spurious association rules (or ghost rules) which
was an undesirable side effect of the process. A complete description of the
various methods for data distortion for association rule hiding may be found in
[106]. Another interesting piece of work which balances privacy and disclosure
concerns of sanitized rules may be found in [89].
The broad idea of blocking was proposed in [20]. The attractiveness of the
blocking approach is that it maintains the truthfulness of the underlying data,
since it replaces a value with an unknown (often represented by '?') rather
than a false value. Some interesting algorithms for using blocking for associa-
tion rule hiding are presented in [97]. The work has been further extended in
[96] with a discussion of the effectiveness of reconstructing the hidden rules.
Another interesting set of techniques for association rule hiding with limited
side effects is discussed in [113]. The objective of this method is to reduce the
loss of non-sensitive rules, or the creation of ghost rules during the rule hiding
process.
In [6], it has been discussed how blocking techniques for hiding association
rules can be used to prevent discovery of sensitive entries in the data set by an
adversary. In this case, certain entries in the data are classified as sensitive,
and only rules which disclose such entries are hidden. An ecient depth-first
association mining algorithm is proposed for this task [6]. It has been shown
that the methods can effectively reduce the disclosure of sensitive entries with
the use of such a hiding process.
5.2 Downgrading Classifier Effectiveness
An important privacy-sensitive application is that of classification, in which
the results of a classification application may be sensitive information for the
owner of a data set. Therefore the issue is to modify the data in such a way
that the accuracy of the classification process is reduced, while retaining the
utility of the data for other kinds of applications. A number of techniques have
been discussed in [21, 83] in reducing the classifier effectiveness in context of
classification rule and decision tree applications. The notion of
parsimonious
downgrading
is proposed [21] in the context of blocking out inference channels
for classification purposes while mining the effect to the overall utility. A
system called Rational Downgrader [83] was designed with the use of these
principles.
The methods for association rule hiding can also be generalized to rule
based classifiers. This is because rule based classifiers often use association rule
mining methods as subroutines, so that the rules with the class labels in their
consequent are used for classification purposes. For a classifier downgrading
approach, such rules are sensitive rules, whereas all other rules (with non-class
attributes in the consequent) are non-sensitive rules. An example of a method
