Self Adaptive Methods for evolutionary Based Sensitive Rule Hiding

Abstract

Privacy-Preserving Data Mining (PPDM) protects users privacy by using Association Rule Hiding (ARH) technique. An example of an ARH approach is that data distortion involves replacing one attribute value with another in a subset of transactions to conceal private association rules. For ARH with less difficulty in the transaction database, the Cuckoo Optimization Algorithm for Association Rule Hiding (COA for ARH) was proposed. Each cuckoo in COA for ARH used a data distortion technique to obscure crucial information in the transaction database. However, each cuckoo in COA for ARH only altered a small percentage of the original database to clean it up. It is insufficient for the numerous data types. Furthermore, COA for ARH still has a sizable off-target effect on insensitive rules. In the first stage of the research the original dataset is pre-processed so that a set of transactions supports the sensitive rules. As a result, the COA method can be used for a wide range of data. A novel fitness function was proposed with several targets to improve the preservation capabilities of the COA for the ARH algorithm. Pareto-optimal solution for association rule hiding has been presented, which uses various objectives to get the best possible collection of solutions. Crowding Distance is a solution for determining the best possible solutions based on their Pareto-optimal performance. For this reason, the Improved COA for ARH-CD (ICOA for ARH- CD) works with many datasets and conceals sensitive rules with minimal effort. newlineThe second phase of the research proposes Quality Preserving ICOA for ARH (QP-ICOA for ARH) to conceal the sensitive rules using many LHS and RHS transforms. A rule s sensitivity and the item s non-sensitivity are correlated, and the item with the highest correlation is selected. Each cuckoo transaction includes a handpicked- out component that is periodically eliminated and replaced based on objective standards and sensitivity.