Sensor network data fault detection with maximum a posteriori selection and bayesian modeling

Abstract

Current sensor networks experience many faults that hamper the ability of scientists to draw significant inferences. We develop a method to systematically identify when these faults occur so that proper corrective action can be taken. We propose an adaptable modular framework that can utilize different modeling methods and approaches to identifying trustworthy sensors. We focus on using hierarchical Bayesian space-time (HBST) modeling to model the phenomenon of interest, and use maximum a posteriors selection to identify a set of trustworthy sensors. Compared to an analogous linear autoregressive system, we achieve excellent fault detection when the HBST model accurately represents the phenomenon.

A Framework and Classification for Fault Detection Approaches in Wireless Sensor Networks with an Energy Efficiency Perspective

“…Ni and Pottie [17] design a two-phase modular fault detection framework which includes four modules: blind modeling, trusted sensor selection, model reevaluation, and sensor evaluation. They first use prior knowledge of the phenomenon behavior to determine the parameters of the Hierarchical Bayesian Space Time (HBST) model adopted for sensor data modeling.…”

A Survey of Research on Data Corruption in Cyber–Physical Critical Infrastructure Systems

“…Ni and Pottie [48] propose using a hierarchical Bayesian space-time model to detect trustworthy sensors. The disadvantage of this technique is the amount of work required to set up the model.…”

An error-free data collection method exploiting hierarchical physical models of wireless sensor networks

“…Moreover, it ignores the possible breakdowns in the correlation between neighbouring sensor readings, which may result in wrong validations. Ni [11] proposes Hierarchical Bayesian SpaceTime (HBST) modelling to find faulty data. Compared with linear autoregressive system, this work has a similar detection rate but lower false-positive rate.…”

“…The validation rule is stated as Eqn. (11), where xi,t denotes a data entry from sensor i at epoch t, and boolj(xi,t) is the verification result from node j.…”