Distributed fault-tolerant event detection and faulty sensor detection in wireless sensor networks

In this dissertation, we solve two applications in wireless sensor networks: distributed fault-tolerant event detection and distributed faulty sensor detection. In event detection, we consider two important problems for distributed fault-tolerant detection in wireless sensor networks: (1) how to address both the noise-related measurement error and sensor fault simultaneously in fault-tolerant detection? (2) how to choose a proper neighborhood size n for a sensor node in fault correction such that the energy could be conserved? We propose a fault-tolerant detection scheme that explicitly introduces the sensor fault probability into the optimal event detection process. We mathematically show that the optimal detection error decreases exponentially with the increase of the neighborhood size. Experiments with both Bayesian and Neyman-Pearson approaches in simulated sensor networks demonstrate that the proposed algorithm is able to achieve better detection and better balance between detection accuracy and energy usage. Our work makes it possible to perform energy-efficient fault-tolerant detection in a wireless sensor network. We further extend the work to the adaptive scheme by considering event boundary conditions.;In wireless sensor networks, faulty sensors may produce incorrect data and transmit the data to other sensors. They may cause inappropriate data fusion. Furthermore, they would consume the limited energy and bandwidth of sensor networks. We propose a distributed faulty sensor detection scheme and restrict our assumption to that the sensor fault probability or reliability is unknown and data to be sensed has Gaussian distribution with unknown parameters. Each sensor obtains a global convergency data through data fusion and makes a local 3-level decision by hypothesis testing against the global convergency data. Unlike the previous works, the detection threshold is independent on the sensor reliability. A final decision about the sensor is obtained by fusing the decisions of its neighbors. Each sensor only communicates with its neighbors in the detection process. Experiment results demonstrate that the proposed algorithm in sensor faulty detection is able to achieve better detection accuracy even without knowledge of senor reliability and parameters of data distribution. For sensor data with large variance, the proposed algorithm still provides better performance.;A model based process security information prediction with sensors under concept of security information completeness is also proposed. To predict system runaway or critical security situation caused by an unknown attack, a methodology to estimate system states and the attacks is thoroughly discussed. The prediction of the system trajectory is obtained based on these estimations. The proposed prediction method can catch the trends of the runaway trajectory, hence being able to quickly and accurately detect a critical security situation. The false alarm and miss rate in prediction are thereafter greatly reduced. It provides a practical method for chemical process security.