Anomaly Detection And Data Cluster In Wireless Sensor Networks

Advances in wireless communication, sensor technique and embedded computing have enabled the deployment of lower cost and more efficient wireless sensor network (WSN) in many areas, such as national military, national security, environment monitoring, traffic management, and medical health. WSNs constitute a new paradigm of ambient monitoring with many potential applications. Typically formed by thousands of nodes of small dimension, they use ad-hoc communication and have scarce resources regarding energy, bandwidth, processing capacity and storage. WSNs are typically designed to gather data in inhospitable places and might be involved in critical applications. However, due to low power energy, limited computing and communication ability of WSN, the problems in WSN are more challenging.Anomaly detection aims at detecting, recording and predicting the abnormal behaviors in WSN, and finally reporting the abnormal warning. We proposed an probability-based approximate algorithm for anomaly detection which is based on rule-based anomaly detection. The algorithm satisfied the probability requirement of user’s input. And we propose some relative proof and relative parameters of the algorithm. The experiment shows that, it is sufficient to reduce the cost of communication band and power energy when choosing appropriate parameters.Due to the scale and density of wireless sensor network, it’s impossible to query the whole sensor data of the network. Cluster is an algorithm which can reduce the data redundancy, and save more energy. We proposed a cluster algorithm based on edge detection. According the edge of the data in network, we can cluster the data surrounding by the edge nodes. The experiments demonstrate that our algorithm is effective to save the energy cost and increase the network lifespan.