An Efficient Approach for Privacy Preserving Distributed Clustering in Semi-honest Model Using Elliptic Curve Cryptography

Abstract

In this paper, we propose an approach that illustrates the application of Elliptic Curve Cryptography (ECC) in Privacy-preserving distributed K-Means Clustering over horizontally partitioned dataset. We believe that the conventional cryptographic approaches and secret sharing schemes for privacy-preserving distributed K-Means clustering, are not scalable due to the higher computational and communication cost. Elliptic Curve based cryptosystems offer much better key size to security ratio in comparison. Hence, we use ECC based ElGamal scheme in distributed K-Means clustering to preserve privacy. Our approach avoids multiple cipher operations at each site and hence is efficient in terms of computational cost. We also achieve a reduction in the communication cost by allowing parties to communicate in a ring topology. Our experimental results show that our approach is scalable in terms of dataset size and number of parties in a distributed scenario. We carry out comparative analysis of our approach with existing approaches to highlight the effectiveness of our approach.