The Data Fusion Technology Based On Wireless Sensor Network (wsn) Research

Wireless sensor networks (WSNs) consists of large number of tiny sensors andthose sensors are of limited communication capability and are deployed in specificregions for target sensing and monitoring in a self-organized manner. Additionally,because of the limited communication range, individual nodes often have overlappingsensing ranges in order to improve the accuracy and correctness of the objectmonitoring. Data fusion technology takes advantage of this feature and can reduce anddelete the redundant data during sensor nodes send their collecting data to the othersensors or base station.Many WSN nodes are deployed in unattended hostile areas, therefore security issuemust be taken in consideration when these networks are being designed. Further, lack ofphysical security makes sensor nodes easy to be compromised by intruders who willlater attack the whole network. Thus, compromised sensor nodes pose a challengingconstraint for the designing of data fusion, and the data fusion algorithm must beenergy-efficient and can operate safely under the existence of compromised nodes.However, many existing data fusion algorithms assume that all the sensors in thenetwork are trusted, thus they are not fit for the security sensitive WSNs. Reputationmechanism can be used to overcome the deficiencies of cryptography in the process ofdata fusion. Through the study of the characteristics of WSNs as well as data fusiontechnology, a trustful data fusion mechanism based on reputation mechanism isproposed in this dissertation, which can mitigate the effect of compromised nodes on thenetwork. Different from the general reputation that computes all the actions of a sensornode, the proposed reputation mechanism can prevent a compromised node fromcovering its bad action regarding certain task by behaving well for another one.Reputation of fusion task is used in this paper to select data aggregator, and reputationof sensing task is used to weigh the data reported by sensor nodes to the data fusionpoint. Simulations show that the proposed method can increase the reliability of datafusion, and using task-based reputation is more effective than using general reputationwhen evaluating the reputation of a sensor node. In addition, for mission critical wireless sensor networks, not only the energyconsumption of sensor nodes but also the correctness of the data fusion results is critical.This dissertation proposes a fault tolerant data fusion scheme that can eliminate the falsedata sent by malfunctioning or compromised sensor nodes. To conserve energy whileeliminating false data, an in-network outlier detection technique that is based onLocality Sensitive Hashing scheme is used. Simulation results show that the proposedmethod can reduce the number of false data transmissions and thus can increase theaccuracy of data fusion.