Privacy Preserving C4.5 Algorithm Over Horizontally Partitioned Data

Abstract

Privacy preserving decision tree classification algorithm is to solve such a distributed computation problem that the participant parties jointly build a decision tree over the data set distributed among them, and they do not want their private sensitive data to be revealed to others during the tree-building process. The existing privacy preserving decision tree classification algorithms over the data set horizontally partitioned and distributed among different parties only can cope with the data with discrete attribute values. This paper propose a solution to privacy preserving C4.5 algorithm based on secure multi-party computation techniques, which can securely build a decision tree over the horizontally partitioned data with both discrete and continuous attribute values. Moreover, we propose a secure two-party bubble sort algorithm to solve the privacy preserving sort problem in our solution