Sequential Fusion Estimation for Sensor Networks With Deceptive Attacks

We consider multisensor fusion estimation for sensor networks with deceptive attacks. The approach of sequential fusion is applied to sequentially detect attacks and thereafter fuse attacked data from different sensors to generate optimal state estimates for linear dynamic systems and reveal the relations between the attack rate and the estimation performance. The idea behind the novel approach is that by exploring useful portion of data from different sensors, the integrity of a useful data for accurate state estimation may still be preserved through multisensor fusion even after being attacked. First, a design method is presented for the sequential measurement fusion estimator with attack detections and some useful relations are established between the parameters characterizing the attacks and the measurement fusion estimation performance. Then, a sequential state fusion estimator with attack detections is designed and a compensation strategy is proposed to alleviate the performance degradation caused by attacks. Both simulations and experiments of a target tracking system are presented to demonstrate the effectiveness of the proposed results.