Abstract—Distributed computing demands training methods that handle distributed input data. While training, as the parties that collaborate are concerned about the privacy of their data, the concept of privacy preservation should be extended in data mining classifiers. In this paper, data holders make practical use of their data to construct a precise classifier model by not revealing either their training data or the intermediate results. We propose a privacy preserving two-party Naive Bayes classifier for horizontally partitioned distributed data. This protocol is built such that both the parties through their random shares compute probabilities, mean and variance. To classify a new instance with numeric attributes, parties need to jointly cooperate with each other. The correctness and the security analysis of our algorithm are provided.
Index Terms—Privacy preserving, data mining, horizontal partitioned, paillier cryptosystem, RSA cryptosystem, secure computations, homomorphic property, commutative property.
K. S. Hareesha is with the Department of Computer Applications, MIT, Manipal, India (e-mail:hareesh.ks@manipal.edu).
M. Sumana is with M. S. Ramaiah Institute of Technology, Bangalore, India (e-mail: sumana.m@msrit.edu).
[PDF]
Cite:K. S. Hareesha and M. Sumana, "Secure Two Party Privacy Preserving Classification Using Encryption," International Journal of Information and Electronics Engineering vol. 6, no. 2, pp. 67-71, 2016.