A Metric-correlation-based Fault Detection Approach In Wireless Sensor Networks

Wireless sensor networks(WSNs)have received significant attentions in recent years,and can provide effective and economically viable solutions for a large variety of applications,such as military operations,environmental surveillance and intelligent transportation,etc.However,WSNs consist of a large number of low-cost and low-powered sensor nodes,deployed in uncontrolled or even harsh or hostile environments,which make them error-prone,have unsatisfactory reliability,and encounter various faults and failures during their operations.The nature of WSNs entails that fault detection is vitally important to enhance the efficiency,reliability and applicability of WSNs.In recent years,many detection approaches relying on specific rules or inference models,requiring a priori knowledge of network faults,have been proposed to distinguish faulty sensor nodes by exploring spatial-temporal correlations among sensor readings.However,many faulty sensor nodes that may not generate anomalous sensor readings,and potential failures with unknown types and symptoms remain undetected.In this paper,we propose a metric-correlation-based fault detection approach in wireless sensor networks.It is motivated by the fact that the system metric correlations of most fault-free sensor nodes usually show strong similarities or perform regularly in successive time windows,whereas different patterns of such correlations indicate potential failures.Our approach explores internal status inside sensor nodes using correlation value views,and detects fault sensor nodes in time and spatial domain respectively.In aspect of time domain detection,an improved cumulative summation(CUSUM)algorithm is used to track gradual changes or abrupt changes.Once any changes occur in correlation value time sequences,potential and highly system-related failures can be detected.The apply of metric correlations has made our approach with high energy efficiency and low computational complexity,since no communication overhead is incurred and CUSUM algorithm is simple for computation.In aspect of spatial domain detection,an improved Neighbor-based Local Density Clustering Analysis(NLDCA)algorithm based on the Neighbor-based Local Density Factor(NLDF)is applied to cluster similar correlation value views together,thus potential faulty sensor nodes with abnormal views not belonging to any cluster can be detected.The introduction of NLDF contributes to the high clustering accuracy and time efficiency of NLDCA algorithm.Finally,we provide the simulation to compare our approach with existing fault detection approaches.Simulation results demonstrate our approach performs well in respects of higher detection accuracy and lower false positive rate even under high node failure ratios and dense distribution conditions.