Privacy-preserving distributed mining of association rules using Elliptic-curve cryptosystem and Shamir’s secret sharing scheme

TLDR: This work proposes two protocols, which are securely generating global association rules in horizontally distributed databases, and incorporates Shamir’s secret sharing scheme in the second protocol, which provides privacy by preventing colluding sites and external adversary attack.

Abstract: Distributed data mining has played a vital role in numerous application domains. However, it is widely observed that data mining may pose a privacy threat to individual’s sensitive information. To address privacy problem in distributed association rule mining (a data mining technique), we propose two protocols, which are securely generating global association rules in horizontally distributed databases. The first protocol uses the notion of Elliptic-curve-based Paillier cryptosystem, which helps in achieving the integrity and authenticity of the messages exchanged among involving sites over the insecure communication channel. It offers privacy of individual site’s information against the involving sites and an external adversary. However, the collusion of two sites may affect the privacy of individuals. To address this problem, we incorporate Shamir’s secret sharing scheme in the second protocol. It provides privacy by preventing colluding sites and external adversary attack. We analyse both protocols in terms of fulfilling the privacy-preserving distributed association rule mining requirements.