The Research On Anomaly Detection Based On Behavior In Wireless Sensor Networks

Wireless sensor networks have strict security requirement in many applications. Due to some limitations of its hardware features, wireless sensor networks face more serious security situation compared with traditional networks. So security problems become an important issue in wireless sensor network research. As a positive and security defense technology, anomaly detection has a lot of achievements in the application of wireless sensor networks, however, it also has some problems such as the complexity of modeling, the huge calculation work, the risk of single point, etc.This article talks about some commonly used anomaly detection technologies in wireless sensor networks through a sensor node's behavior. The research work can be classified as follows:First, this paper gives a general overview of the current research about WSN, including system structure, features and applied areas. Based on this, it summarizes the security technologies, and compares those available abnormal detection technologies.Secondly, the location-based anomaly detection method largely depends on the available deployment information or additional hardware. Moreover, many methods adopt centralized detection by base station, which may lead to such problems as the huge communication amount and the invalidation of single point. To solve these problems, we propose an anomaly detection method based on distributed node localization. On top of a clustering WSN architecture, the clustering network can reduce the communication overhead between sensor nodes by utilizing clustered topology, and doesn't have the risk of single point of failure. This method doesn't need any deployment knowledge or hardware. According to compute the location and neighbor list from the sensor node, update its weight by the cluster head, this scheme can filter the anomaly node. Through simulation, we verified the correctness and efficiency of our scheme.Thirdly, we propose an anomaly detection method based on data transmission quality by using nodes' historical transmission statistical data. This method defines a data transmission quality function, which quantized evaluates a node's transmission quality by historical transmission statistical data. The feature is that for normal node it keeps a constant while for dubious node it continuously decreases.The method is simple, in that it is able to determine the exceptions just according to the varying value of DTQ function, by a voting procedure. Simulation result shows that the quantized detection method gives a satisfy detection performance for slight network abnormal.