Event-triggered attack detection and state estimation based on Gaussian mixture model

Abstract

Under the framework of event-triggered transmission mechanism, the problem of attack detection and state estimation of multi-sensor linear time-invariant systems under static attacks is considered. First, for each transmission channel, the sensor collects measurement information according to an event-triggered mechanism to reduce unnecessary energy consumption. Then, inspired by the clustering algorithm in machine learning, a detection mechanism based on Gaussian mixture model, which can set a confidence level for the measurement of each sensor is proposed. Finally, centralised data fusion is performed according to the results of attack detection and event-triggered judgement to realise remote state estimation. A numerical example proves that the proposed algorithm can locate the damaged sensor, save the network transmission bandwidth under the premise of ensuring accuracy and efficiency of sensor estimation.