Privacy Preserving Pattern Discovery in Distributed Time Series

The search for unknown frequent pattern is one of the core activities in many time series data mining processes. In this paper we present an extension of the pattern discovery problem in two directions. First, we assume data to be distributed among various participating peers, and require overhead communication to be minimized. Second, we allow the participating peer to be malicious, which means that we have to address privacy issues. We present three problems along with algorithms to solve them. They are presented in increasing order of complexity according to the extensions we are pursuing, i.e. distribution and privacy constraints. As the main result we present our secure multiparty protocol for the privacy preserving pattern discovery problem.