Privacy-preserving mining of decision trees using data negation approach

With the ever increasing need to share data across organizations, the demand for its privacy is also becoming increasingly important. In the past, several techniques have been proposed for privacy-preserving data mining. In this paper, we propose a privacy-preserving technique to build decision trees. We refer to it as negation technique. Since our technique does not employ any cryptographic techniques, it is computationally efficient. However, as a trade-off, it requires extra storage at the data owner site. We show that there is no loss in accuracy of the resulting decision tree due to the proposed data transformations. In this technique, we convert all the tuples in the database to anti-tuples, and then using the results obtained, we build the original decision tree from the transformed tuples. Privacy can be enhanced further by using cryptographic techniques along with this technique.