Association rules mining in vertically partitioned databases

Privacy concerns have become an important issue in Data Mining. This paper deals with the problem of association rule mining from distributed vertically partitioned data with the goal of preserving the confidentiality of each database. Each site holds some attributes of each transaction, and the sites wish to work together to find globally valid association rules without revealing individual transaction data. This problem occurs, for example, when the same users access several electronic shops purchasing different items in each. We present two algorithms for discovering frequent itemsets and for calculating the confidence of the rules. We then analyze the algorithms privacy properties, and compare them to other published algorithms.