A Robust Statistical Scheme to Monitor Transient Phenomenon in Sensor Networks

Wireless sensor networks have been deployed for various critical monitoring applications in hostile environments such as monitoring the concentration levels of hazardous gas species in a battle field. Due to the sensitivity of such applications, it becomes mandatory to record the transient variations in the phenomenon, and take corrective actions, if necessary. At the same time it is important to shield the network from false data injection by adversaries who intend to disrupt the functioning of the system. While many schemes exist to prevent false data injection, they are counterproductive to preserving the transient observations. We devise a robust statistical scheme to monitor transient phenomenon while being immune to false data injection attacks. The key idea of our scheme is to require each sensor node to report a statistical digest of recent sensed readings in addition to the current reading; then inter-sensor statistical tests are designed and utilized to help preserve transient data while restricting the impact of false data injection significantly. Detailed theoretical analysis and in-depth simulations are presented to corroborate our scheme.