Decision Tree based Data Reconstruction for Privacy Preserving Classification Rule Mining

G. Kalyani, M.V.P. Chandra Sekhara Rao, B. Janakiramaiah


Data sharing among the organizations is a general activity in several areas like business promotion and marketing. Useful and interesting patterns can be identified with data collaboration. But, some of the sensitive patterns that are supposed to be kept private may be disclosed and such disclosure of sensitive patterns may effects the profits of the organizations that own the data. Hence the rules which are sensitive must be concealed prior to sharing the data. Concealing of sensitive patterns can be handled by modifying or reconstructing the database before sharing with others. However, to make the reconstructed database usable for data analysts the utility or usability of the database is to be maximized. Hence, both privacy and usability are to be balanced. A novel method is proposed to conceal the classification rules which are sensitive by reconstructing a new database. Initially, classification rules identified from the database are made accessible to the owner of the data to spot out the sensitive rules that are to be concealed. In the next, from the non-sensitive rules of the database, a decision tree will be constructed based on the classifying capability of the rules, from which a new database will be reconstructed. Finally, the released reconstructed database to the analysts reveals only non-sensitive classification rules. Empirical studies proved that the proposed algorithm preserves the privacy effectively. In addition to that utility of the classification model on the reconstructed database was also be preserved.

Full Text:


Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.