Abnormal Data Detection Method In Wireless Sensor Network For Environment Monitoring

With the development of environment monitoring technology using wireless sensor network,abnormal data detection has recently drawn more attention from both academic and industry fields. Inorder to monitor the potential emergency (such as forest fire, air pollution, landslides, etc) in time, it iscritical to detect the abnormal data collected by sensor nodes. Therefore, detection of abnormal datacorrectly are of great importance.According to the principles of clustering in data mining, a method of abnormal data detectionbased on DBSCAN is proposed in this paper. The proposed method uses the distance to define thesimilarity of data for the cluster partitioning, and extracts the environmental feature set by the DBSCANalgorithm, and then detects the abnormal data with the environmental feature set. During the detectingprocess, the sensor nodes only transport the abnormal data to sink, which reduces the communicationoverhead between the nodes and sink. When there is an event happened, the proposed method can detectand give alarm in time. The communication among the nodes, gateway and base station is required, so acommunication interface protocol of gateway node is designed.Finally, three experiments are carried out based on an indoor wireless sensor network. For theartificially injected abnormal data and event caused abnormal data, different experiments are designedto verify the effectiveness of the proposed method. Experimental results show that the proposed methodcan detect the abnormal data correctly and real-timely.