Wireless sensor networks (W) have provided a very attractive and fast-growing research fields that have been receiving significant attention in recent years. These networks are applied extensively in military, medicine, industry and home applications. One application involves event detection in different environments, where we want to know whether an event has occurred or not. Generally event can be a change in environment parameters such as temperature, pressure, humidity, etc. These networks are prone to failure and defects as a result of having limited source of energy, using cheap and low quality software or hardware equipments, exposing to harsh and hazardous environments, being compromised by malicious attacks and faulty nodes. Therefore, fault tolerance has been a challenge for event detection researches in wireless sensor networks. Faults may occur due to several reasons. Fault can be referred to a simple temporarily failure of a node due to distributing process, or to a node compromised by enemy and it’s maliciously behavior. Since the failure probabilities of the network elements which are in interaction with environment are more than other elements of the network, the sensing unit of sensor nodes and communication links are investigated more often. Two types of faults are considered in this thesis, symmetric faults and asymmetric or byzantine faults. In symmetric faults, sensor nodes only sense the phenomena incorrectly while the other parts of the node such as processor, memory, radio transceiver and etc work properly and behave quite similar to the other nonfaulty nodes in terms of the rest of features and functions. In byzantine faults, a nonfaulty node could be compromised by an adversary which tend to disturb network functionality. Byzantine node transmits conflicting results or data to other nodes intentionally. The main focus of this thesis is those symmetric faults. Based on coverage range, events are divided into pervasive and non pervasive categories. At the event occurrence instance, all nodes are inside of event region under pervasive event assumption, in contrast, some nodes are out of event region under non pervasive event assumption. In detection algorithms which assume non pervasive events, boundary nodes have serious problems in making correct decision. In this thesis only pervasive events are considered. Symmetric fault tolerant event detection algorithms need to consider the minimum data transmission usage due to restricted sources of energy, and having the low error rate in event detection. This thesis will propose a symmetric fault tolerant event detection algorithm which takes two important criteria into account... Key Words: Wireless sensor network, Event detection, Pervasive and nonpervasive event, Symmetric and Byzantine fault, Fault tolerant
