Outlier Detection in Wireless Sensor Networks Based on Neighbourhood

Wireless sensor networks contain millions of nodes deployed in a spatially dispersed manner. These sensors are low battery powered devices having limited storage and computation power. The data collected by these sensors may be subjected to error due to environmental fluctuations, interference in wireless communication or wearing of sensors with time. These erroneous data deviate significantly from the rest of the data. To solve this issue, we present a new technique named Outlierness Factor based on Neighbourhood to detect and analyse the outliers in sensor network. Proposed detection approach is time efficient and scalable. Further, outlier data are classified as errors due to sensor malfunctioning or actual detected events such as fire detection, weather changes, earthquakes, landslide etc. The capabilities of the proposed approach have been evaluated on real dataset obtained from Intel Berkeley research lab and synthetic datasets. The results show the effectiveness of the proposed approach in contrast to the previously dealt approaches.