Privacy Preserving Attribute Reduction Based on Rough Set

Attribute reduction, as a part of preprocesses, plays an important role in data mining. Privacy ought to be preserved while conducting attribute reduction on distributed data. However, to the best of our knowledge, there exists no algorithm about attribute reduction for the present. In this paper, we represent two privacy preserving attribute reduction algorithms based on rough set. One is on the vertically partitioned data. We develop secure sum of matrices and secure set intersection for it. The other is on the horizontally partitioned data, mainly using secure set union. The correctness and security of the two algorithms are also analyzed. The results show that both of the two algorithms are correct and secure.