The Research Of Intrusion Detection In Wireless Sensor Networks

The gradual matureness of the sensor technology makes it possible to produce thelow-cost, low power consumption, multi-functional micro-sensors in a large scale.Information acquirement from sensors has changed from the unitary in the past tointegration, miniaturization and networking. The wireless sensor networks (WirelessSensor Networks, WSNs) therefore emerges. A WSN is a task-oriented andself-organizing network system, consisting of a large number of micro sensor nodesdeployed in the designated areas by radio communication. Currently, the WSNs havebeen applied into a variety of fields, with the intrusion detection to be an importantapplication. The WSNs-based intrusion detection has developed into a key technologyin the safety network programs. Due to the small size, high-density distribution, storage,computing and transferring data characteristics of sensor nodes, WSNs-based intrusiondetection has the capability to monitor and detect intrusion within the network in time,and transfers the information to the network administrator by wireless communication.This thesis has studied the WSNs-based intrusion detection. It firstly investigatesthe sensor predefined distribution effect on intrusion detection. In the research of theWSNs-based intrusion detection, what kind of predefined distribution can lead to ahighly efficient intrusion detection is a valuable research problem. In this thesis, twodifferent predefined distribution are studied comparatively, i.e., the Gaussiandistribution and the uniform distribution, on the performance of intrusion detection. Byinvestigating the effect of the number of sensor nodes and sensor nodes’ sensing rangeon intrusion detection distance in wireless sensor networks, this paper analyzes the twosorts of distribution on the efficiency of intrusion detection. Experimental results showthat, with the increase of the number of sensor nodes or sensing range of sensor nodes,intrusion distance became shorter. In addition, for a small value of sensor number or ashort sensing range of individual sensors, the Gaussian distribution presents a lowerintrusion distance, indicating a better intrusion detection performance. For a large valueof sensor number or a big sensing range, the uniform distribution presents a betterintrusion detection capability. Futhermore, this thesis considers the effect of the discretedegree indicated by the standard deviation of the Gaussian distribution on the intrusiondistance. The results show that, at a certain value of the standard deviation of the Gaussian distribution, the intrusion distance can approach an optimal peak value.Secondly, this thesis has studied what kind of network deployment model is moreefficient on the network coverage in a fixed monitoring region. While the sensor nodesare deployed in the specified location by precise delivery method, this paper researchesthe effects of the number of sensor nodes and sensor nodes’ sensing range on theno-overlap and full network coverage in wireless sensor networks deployed with the2Dcongruent triangle, square and hexagon network configuration. The results show thatcongruent triangle model is the optimal grid model in network coverage.