Event‐triggered differential private distributed fusion estimation for sensor networks with unknown cross‐correlations

Abstract

Recently, preserving privacy and reducing energy consumption become important concerns of sensor networks. However, existing privacy‐preserving estimation methods cause many needless data transmissions, which results in energy waste. Besides, these methods generates fusion estimates with heavy computational burdens, due to needing calculating cross‐covariance matrices of each pair of local estimation errors. Worse still, these cross‐covariances are often unavailable in actual sensor networks. In view of these problems, a novel framework of event‐triggered (ET) differential private distributed fusion estimation is proposed. Within this framework, an ET mechanism is designed to schedule transmissions of local estimates for reducing communication consumptions. Then, with using local perturbation mechanism, conditions for achieving differential privacy and dealing with eavesdroppers which can eavesdrop and fuse local estimates are provided. Based on covariance intersection fusion rule, an ET differential private fusion method is developed, which can guarantee fusion results uniformly stable in spite of completely unknown cross‐covariances. Finally, simulation results verify that the proposed method can preserve privacy and reduce communication consumptions without using cross‐covariances, at the cost of only slight decline of estimation performance.