A New Model for Privacy Preserving Multiparty Collaborative Data Mining

Due to the increasing use of internet, Ecommerce has become part and parcel of everyone's life. So there is a need to understand buying trends of people by the Sellers to improve business. One way to accomplish this is it to analyze the transactions to find out the items that are frequently bought together. In this paper, we propose a model where in multiple parties come together to analyze the buying trends by sharing their individual transactional databases, which we call it as multiparty collaborative data mining. But the privacy of sensitive data in multiparty collaborative mining is a major issue. The participants collaboratively conduct data mining to extract useful information from the database. In this approach, each participant has sensitive and non-sensitive data in their local database. Therefore, an important challenge of privacy preserving collaborative data mining (PPCDM) is how multiple parties efficiently conduct data mining without exposing each participant's sensitive information. This paper proposes the idea of preserving privacy of sensitive information by hiding sensitive information that are in the form of frequent itemsets. As a first step, we find the frequent itemsets using Apriori algorithm. The sensitive frequent itemsets are removed from the database using hiding algorithm and using border revision technique and Binary Integer Programming (BIP) technique we sanitize the database which we finally share with other participants for data mining. We use CPLEX solver to find solutions for hiding sensitive frequent itemsets, which provide solution to the investigated problem of disclosure of sensitive data.