HEMIN: A Cryptographic Approach for Private k-NN Classification

Data mining is frequently obstructed by privacy concerns. In many cases, data is shared with third party for the mining purpose. However, sharing of data for analysis is not possible due to prevailing privacy laws and/or policies. Privacy preserving data mining techniques have been developed to address this issue by providing mechanisms to mine the data while guaranteing certain levels of privacy. In this paper, we address the issue of privacy preserving nearest neighbor search, which forms the basis of many data mining applications. We propose a scheme HEMIN for sharing the data to be mined by way of homomorphic encryption scheme at the data owner end. The nearest neighbors of a given data point is computed while the data point as well as the data points in training set are in encrypted form. This is being done using the third party which is assumed to be semi honest party. The efficiency of the model has been shown over three datasets namely, Iris, German credit database and Australian credit dataset. The misclassification ratio has been computed to show the efficiency of the k-NN classifier. The results show the effectiveness of the proposed method and suggests its applicability for various parametric data mining algorithms.