Privacy-Preserving Single Layer Perceptron

In this paper, we will introduce several notions of privacy, such as k-anonymity, p-indistinguishability, (c,t)-isolation, and SMC-based measures. We then turn our attention to privacy-preserving classification algorithms. We propose two approaches to classify data using a neural-network classifier called perceptron. The first approach is based on probabilistic encryption, while the second one is based on simple random computations. The two approaches are then compared, and the advantages and disadvantages of each are considered.

Keywords:

Privacy-Preserving Data Mining , k-Anonymity , p-Indistinguishability , (c , t)-Isolation , Secure-Multiparty Computation , Perceptron