Random Walk on Random Graph based Outlier detection in wireless sensor networks

Wireless Sensor Network (WSN) is characterized with limited battery power and limited computation capability. Sensor nodes which produce a data set that is different from their counter parts are called Outlier nodes. For example, in a particular room, if we expect temperature of 25°C and if we receive a temperature reading of 73°C. This is called outlier data and the node which produces this data is called outliers nodes. We are considering a system where user is interested only in outlier data. We assumed that outlier data is generated due to the ambiance parameter. We also assumed that instead of sending all the sensed data, nodes are required to send the outlier data in response to a query from the base station. In this paper we found that using traditional routing protocol for communications is not an optimum solution. Traditional routing protocols consume more memory and battery power in the route-request process. We have also suggested that use of "Random Walk on Random Graph Technique" reduces the overhead of transmitting packets. We have proved this with analysis and simulation. We argue that with random walk on random graph mechanism, energy consumption is minimized and numbers of packets flooded in the network is very less.