Deception in wireless sensor networks- predicting intruder behavior

Abstract

Wireless sensor networks (WSNs) have been used in different sectors such as transportation, agriculture, military etc., Since WSNs are generally used in unmanned territories, providing security is an important requirement. Deception in WSNs is a method for providing a security framework. Time-series prediction is a component used to validate the effectiveness of the deception framework employed. The time-series prediction implemented in this thesis requires little memory and computation power to predict within a certain accuracy threshold. It works by scanning the data towards the end of the existing time series, checking for patterns and trends. Averaging is applied if it is not possible to arrive at a definite conclusion based on the data scanned. The algorithm was tested with data obtained from a sink-hole attack in WSNs. Simulation results show that the prediction was most accurate with a look- back value of 10 and input lengths set to 100. The proposed approach requires little memory and computational power and is therefore suitable for WSNs.