Data Validation Algorithm for Wireless Sensor Networks

This paper presents a novel data validation algorithm for wireless sensor network. We applied qualitative methods such as heuristic rule, temporal correlation, spatial correlation, Chauvenet’s criterion, and modified -score as algorithms for validating sensor data samples for faults. Performance of the algorithms is evaluated using real data samples of WSNs prototype for environment monitoring injected with different types of data faults such as out-of-range faults, struck-at faults, and outliers and spike faults. Results show heuristic rule, temporal correlation, spatial correlation, chauvenet’s criterion, and modified -score method sit at different point on accuracy, no single method is perfect in detecting different types of data faults and reports false positives when sensor data samples contain different types of data faults. Selected effective methods such as heuristic rule, temporal correlation, and modified -score are applied successively to data set for detecting different types of data faults but report false positives due to masking effects and increased fault rate. Finally we propose a novel data validation algorithm that uses novel approach in applying heuristic rule, temporal correlation, and modified -score to data set for detecting different types of data faults. Compared to other methods, the proposed novel data validation algorithm is effective in detecting different types of data faults and reports high fault detection rate by eliminating false positives. 1. Introduction Wireless sensor networks (WSNs) include sensor nodes from few to several hundred that can be deployed in remote distributed geographical environment to sense phenomenon and transmit it to base station for performing scientific studies, analysis, and decision making. WSNs has unlimited potential for creating revolution in the area of environment management, industrial process automation, transportation, crisis management, precision agriculture, medical care, defense surveillance, smart buildings and smart cities. Many real time deployments [1–4] show that data samples collected from WSNs are prone to be faulty due to internal and external influences, such as environment effects, limitations of resources, power problems, hardware malfunctions, software problems, and network problems, security attacks, and [5–7]. In Ni et al. [8, Table v] the anothers specify different types of data faults and their possible causes. Here we summarize data fault types that we consider in this paper.(i)Out-of-range faults: sensor data samples that deviate significantly from expected range