Trust based data prediction, aggregation and reconstruction using compressed sensing for clustered wireless sensor networks

Abstract

Sensing and relaying are the primary tasks of sensor nodes in a Wireless Sensor Network (WSN). Hence the recent research focus has been to devise secure, energy efficient ways to predict, aggregate and recover sensor data. In this paper, a novel method for secure data prediction in WSN has been proposed by using a Time Series Trust Model (TSTM) based on Toeplitz matrix and a Trust based Auto Regressive (TAR) process. The impact of the proposed trust model in data prediction and Compressed Sensing (CS) based aggregation and reconstruction is validated using various performance metrics and different attack models. The TAR model for prediction is evaluated against three different attack models. The proposed TSTM model outperformed existing trust model for varying percentage of compromised nodes. TSTM based data reconstruction using the Basis Pursuit (BP) algorithm registers best performance when the percentage of compromised nodes varies between 10% and 40% due to bad mouthing attack.