Predicate-tree based pretty good privacy of data

Abstract

Growth of Internet has led to exponential rise in data communication over the World Wide Web. Several applications and entities such as online banking transactions, stock trading, e-commerce Web sites, etc. are at a constant risk of eavesdropping and hacking. Hence, security of data is of prime concern. Recently, vertical data have gained lot of focus because of their significant performance benefits over horizontal data in various data mining applications. In our current work, we propose a Predicate-Tree based solution for protection of data. Predicate-Trees or pTrees are compressed, data-mining-ready, vertical data structures and have been used in a plethora of data-mining research areas such as spatial association rule mining, text clustering, closed k-nearest neighbor classification, etc. We show how for data mining purposes, the scrambled pTrees would be unrevealing of the raw data to anyone except for the authorized person issuing a data mining request. In addition, we propose several techniques which come along as a benefit of using vertical pTrees. To the best of our knowledge, our approach is novel and provides sufficient speed and protection level for an effective data security. © 2012 IFIP International Federation for Information Processing.