Reputation-Enabled Self-Modification for Target Sensing in Wireless Sensor Networks

Wireless sensor networks provide new tools for sensing physical environments. However, the general existence of faulty sensor measurements in networks will cause degradation of the network service quality and huge burden of the precious energy. While cryptography-based approaches are helpless of information generation, reputation systems are demonstrated of positive results. In this paper, we investigate the benefits of a distributed reputation system in target localization. A node reputation is defined as its measurement performance and is computed by the Dirichlet distribution. By assuming the sensing model of each node to be mixed Gaussian, we use reputation to estimate parameters of the sensing model and modify a node's original measurement. We also develop a reputation-based local voting algorithm to filter the untrustworthy data and then estimate the target location by a particle swarm optimization algorithm. To assure energy efficiency of the proposed approach, we use a reputation-based model to indicate the information importance of each data packet and ensure that a more important packet can be delivered with higher reliability. Finally, we experimentally evaluate the reputation system and demonstrate its accuracy and energy efficiency.