A Node-Oriented Self-Monitoring And Anomaly Detection Mechanism In Wireless Sensor Networks

Wireless sensor networks is compose of wireless sensor nodes which deploy in the environment without any network topology, it can support stars and peer to peer models, the nodes can join or quit freely. Since the features of self-organization, self-management, data centric in WSN, it has broad application prospects in environmental monitoring, intelligent monitoring, data collection, security, military and so on. But limited by the energy, the computing resource, the changeful topology and the multiple-hop data transmission, WSN faces a variety of special attacks such as data interception, Sybil attacks, wormhole attacks and selective forwarding attacks except traditional attacks. Wireless sensor networks have been widely researched both at home and abroad, and the security in WSN is especially important. From analysis of the solution to this issue, we found that combining the WSN nodes monitoring and compromised node detecting together is an effective way. By monitoring each node, we can know the physical information, energy status and behavior of data communication, which can grasp the living condition of the node to help the sink node to balance the network load, extending the living time of the nodes and the network. What's more, by statistically analyzing node communication behavior we can effectively detect compromised nodes in the network, while the attack action is launched by those nodes. Based on the analysis of the existing monitoring model and security algorithms, this paper proposes a new monitoring model to monitor nodes in WSN. Also we define six rules to detect compromised nodes while monitoring the communications between nodes. This thesis focuses on two aspects of WSN self-monitoring and compromised nodes detecting as follow, and achieve good performance.1. Energy consuming is a great problem in WSN monitoring, to this issues we propose geometry-circle model which focus on the cost of building a monitoring model in WSN. This model analyses the connectivity of nodes and construct corresponding geometric model to calculate the number of needed monitoring nodes, and then deploy those nodes in WSN. After the first step, there could be multi-monitoring problem for the uncertainty of network topology and node communication range. So we propose two kinds of optimization algorithms reducing the number of monitoring nodes by analyze the location information of the monitoring nodes and the monitored nodes, and aim to reducing the energy consuming for the further step.2. Defines six rules for detecting compromised nodes in geometry-circle model, to defense the attacks such as Sybil attacks, wormhole attacks and selective forwarding attacks in WSN, using rule-based detection to detect and located compromised nodes. Those rules are made according to the feature and the condition precedent of different attacks, and can be combined to detect compromised nodes, which will effectively avoid the cooperated-deceive and improving the accuracy of detection.