Design of Secure Multi Party Computational Protocols for Classification Rule Mining

Abstract

Use of technology for data collection and analysis has seen an amazing growth in the last couple of decades. Nowadays privacy is becoming an increasingly important issue in many data mining applications. The data mining community has responded to this challenge by developing a new breed of algorithms that are privacy preserving. Specifically, cryptographic techniques outline the set of privacy preserving data mining algorithms for distributed computation environments. However, these algorithms require all parties in the distributed system to follow a massive privacy model and also make strong assumptions about the behavior of participating parties. These conditions do not necessarily hold true in practice. Therefore, most of the existing work in privacy preserving distributed data mining fails to serve the purpose when applied to real-world distributed data mining applications. In this thesis, the novel protocols have been developed for privacy preserving classification rule mining in a distributed data environment. The protocols allow parties to perform mining at their own sites, and transfer their intermediate results only. In this thesis, some Secure Multi-Party Computational (SMC) protocols are provided to protect the privacy of participating parties while still allowing users to mine useful trends and knowledge. Since the developed protocols are not required to transfer actual data, so the protocols are time efficient, memory efficient, secure and real-world adaptable. newlineIn this thesis, SMC protocols designed for privacy preserving classification rule mining in distributed environment are based on ID3 and Naïve Bayes classification algorithm. In the development of SMC protocols, two assumptions are made. In first, all participating parties communicate their intermediate results to each other and in second, all participating parties communicate their intermediate results to Un-trusted Third Party (UTP).