Approach to Privacy Preserved Data Mining in Distributed Systems

Nowadays using data mining techniques is very important for many modern industrial systems. They could help to decide different tasks in these systems related with decision making, forecasting, detection of anomalies etc. More often industrial systems are implemented in distributed way, which leads to necessity of using data mining techniques for distributed data. It could influence the violation of information security. Source data can be disclosed during its transmission through the public networks or in case of capturing data storages. This paper is devoted to solving the information security problem for distributed data mining systems. We present a privacy-preserving DBSCAN clustering algorithm over vertically partitioned data, analyze the security of this algorithm in the context of different adversaries and demonstrate its performance. The suggested approach could be widely used for data privacy preserving in industrial systems.