The Research On Energy-Efficient Intrusion Detection In Wireless Sensor Networks

Wireless integrated sensor networks, which include collecting, processing data and communication, are used more and more widely for its low cost and convenient deployment. Nowadays the researchs of sensor networks are fairly active. The security is one of the key questions in sensor networks. Intrusion detection is a kind of network security technologies used to detect any behavior that will damage or attempt to damage system confidentiality, integrality or availability, and it can provide the reasonable supplement to intrusion prevention mechanism, and construct a second wall of defense for network and system. This thesis mainly focuses on the energy-efficient intrusion detection technology.First, the thesis firstly introduces the security system and the security request in sensor networks, and summarizes the attacks and the invasions which every protocol layers of sensor networks are suffered from. The thesis contrasts exist several kinds of the intrusion detection models in sensor networks, and proposes the design goals of the intrusion detection model.Second, according to the characteristics of sensor network and the specility of the invasions in sensor network, this thesis presents a intrusion detection model based on statistics anomaly in sensor networks. The algorithm establishs models for the normal state of the nodes, and makes decisions through the deviation degree of observed value. The algorithm is fault-tolerant for non-invasion anomaly when the communication between nodes break down or the accident wrongly create anomaly.Third, node replication attacks are harmful in sensor networks, and the existing algorithms are heavy in expenses. This thesis presents a clustering-based algorithm to detect node replication attacks. Through witness nodes recording positional information of sensor nodes, the scheme find replication nodes from any of region, when it executes distributed detection in the cluster region and centralized detection in cluster head.Fourth, in order to verify the validity of the algorithms, we propose a network simulation model on which the algorithms are simulated and evaluated. Results in simulation experiments show that the intrusion detection model based on statistics anomaly can work perfectly and its algorithm is highly effective for the energy conservation; as well as the clustering-based detetion algorithm has higher detection rate and lower energy consumption with the other comparisons.