Contact at mumbai.academics@gmail.com or 8097636691
Responsive Ads Here

Wednesday, 7 February 2018

Secure Mining of Association Rules in Horizontally Distributed Databases

Secure Mining of Association Rules in Horizontally Distributed Databases


We propose a protocol for secure mining of association rules in horizontally distributed databases. The current leading protocol is that of Kantarcioglu and Clifton [18]. Our protocol, like theirs, is based on the Fast Distributed Mining (FDM) algorithm of Cheung et al. [8], which is an unsecured distributed version of the Apriori algorithm. The main ingredients in our protocol are two novel secure multi-party algorithms — one that computes the union of private subsets that each of the interacting players hold, and another that tests the inclusion of an element held by one player in a subset held by another. Our protocol offers enhanced privacy with respect to the protocol in [18]. In addition, it is simpler and is significantly more efficient in terms of communication rounds, communication cost and computational cost. Secure Mining of Association Rules in HDD
CONCLUSION
We proposed a protocol for secure mining of association rules in horizontally distributed databases that improves sig- nificantly upon the current leading protocol [18] in terms of privacy and efficiency. One of the main ingredients in our proposed protocol is a novel secure multi-party protocol for computing the union (or intersection) of private subsets that each of the interacting players hold. Another ingredient is a protocol that tests the inclusion of an element held by one player in a subset held by another. Those protocols exploit the fact that the underlying problem is of interest only when the number of players is greater than two.
One research problem that this study suggests was described in Section 3; namely, to devise an efficient protocol for inequality verifications that uses the existence of a semi- honest third party. Such a protocol might enable to further improve upon the communication and computational costs of the second and third stages of the protocol of [18], as described in Sections 3 and 4. Other research problems that this study suggests is the implementation of the techniques presented here to the problem of distributed association rule mining in the vertical setting [31], [33], the problem of mining generalized association rules [27], and the problem of subgroup discovery in horizontally partitioned data [16]

No comments:

Post a Comment