Effective Privacy Preservation over Composite Events with Markov Correlations

With the rapid development of radio frequency identification (RFID) and sensor networks, complex event processing (CEP) has attracted extensive attention. Meanwhile, privacy preservation techniques towards composite events are becoming increasingly important. However, practical applications may generate a large number of uncertain data due to inaccurate reading and missing reading. An effective method to model such uncertain data is leveraging the Markov model while conducting CEP over it may cause severe privacy leakage. In this paper, we focus on the privacy preservation on the composite events modeled by Markov chains. Due to the inherent uncertainty and correlations of the data, traditional techniques for privacy preservation fail to support the problem. Specifically, we propose two methods (Type_S and Instance_S) with different optimization goals in processing efficiency and the published result amount. The empirical evaluation verifies the efficiency and applicability of our proposed methods.