Research Of Intrusion Detection For Wireless Sensor Networks

Intrusion detection is an important field in security of wireless sensor networks. Therefore the research of intrusion detection for wireless sensor networks has very important theoretical and practical significance.Considering the constraints and features of wireless sensor networks, we propose a data-oriented intrusion detection model for wireless sensor networks. The largest application of wireless sensor networks is to monitor the environment data, so it has the data-centric feature. Otherwise, as sensor nodes are limited in power, computational capacities, and memory, if the complexity of modeling or the requirement of the sensor node is high, the function of the intrusion detection system will be infeasible in practical application. Therefore, the intrusion detection model should fits the characteristics of wireless sensor network and lower requirement of the sensor node under the premise of guaranteeing the accuracy of intrusion detection.After analysis these characteristics, the paper uses hierarchical architecture, let the intrusion detection programs run on the task management node to reduce the sensor nodes’limits in power, computational capacities, and memory. The intrusion detection model is divided into passive detection and active detection, combining the two ways to detect intrusion behavior. When the passive detection is abnormal, then start the active detection. This method improves the detection accuracy, and not to increase the burden of network communication. Passive detection predicts the future data by ARMA model, using sensory data of time correlation. It uses the historical data collected by the sensor nodes to establish model and analyze, then forecast the future data, and compares the actual and predicted values. If the deviation is without the allowable range of the error threshold, then we thought the actual data is abnormal. This method makes full use of the data-centric characteristic of the wireless sensor networks. Then active detection judges the node status further to conclude the final results.In this intrusion detection system, by exploiting the temporal correlation of data series generated by a node, this paper proposed an ARMA time series model based data aggregation method which is designed to forecasts future sensing data and detect the abnormal behavior.Combining the intrusion detection model, we designed the intrusion detection system based on the system structure, the communication mode and the node performance of wireless sensor networks. There are there models in this intrusion detection system, including management module, data acquisition module, data preprocessing module, intrusion detection module, alarm and display module. The five modules jointly completed a series of functions such as the data collection, intrusion detection and alarm display.Finally, the article build a network for experiment, and detect the spoofing attack using the intrusion detection system. Test results show that the proposed intrusion detection system can detect intrusion effectively.