Research On Intrusion Detection In Wireless Sensor Networks

With the increasing applications of wireless sensor networks (WSNs) in military surveillance, medical treatment and environment monitoring, etc., more and more people regard the security of WSNs as an emergent problem. The study of intrusion detection is a key field in the research of the security of WSNs. At present, it is meaningful in theory and practice to design an intrusion detection scheme that could fit the features of WSNs.Wireless sensor networks are susceptible to many types of attacks because of their constrained resources, such as limitation in computation power, memory and battery lifetime, etc. In this paper, we propose a novel dynamic intrusion detection scheme for cluster-based wireless sensor networks. Our scheme can solve real-time intrusion detection and low power consumption well in WSNs.We will make use of Mahalanobis distance to identify attackers. Multiple attributes of the sensor nodes are taken into consideration in our scheme to identify attackers to minimize the interference of any single attribute to the effect of intrusion detection. We group the nodes in a sensor network into clusters based on the special relationships with a cluster head in each cluster. Cluster heads are in charge of the sensor nodes in the clusters and the communication with the sink node. Our scheme will set different detection parameters for different clusters due to the differences of clusters to achieve a more appropriate detection algorithm based on network conditions. For a cluster, the network condition is different in different periods. The scheme will also adjust the parameters in the same cluster in accordance with the network conditions.In order to finish intrusion detection, our intrusion detection system will include the sink node. A complex hybrid intrusion detection algorithm is used in the sink node since it has enough resources. The sink node is in charge of parameter adjustment in the clusters.We will compare our scheme with other intrusion detection schemes through simulation. All the experiment data is collected from a real sensor network. The experiment shows that our scheme can achieve higher detection accuracy rate than the insider attacker detection scheme while getting a lower false alarm rate.At last, this thesis sums up the research results of our subject and discusses our future research in this field. We point out that research on intrusion detection in wireless sensor networks has a great deal of practical value.