IOT End Node Behavior Analysis

With the development of information technology, Internet of Things (IOT) launch a third wave of information technology, which represents a change in lifestyle, and involves all areas of production and life. However, the development of IOT not yet mature, and sensors security monitoring in the production environment is still a problem in the field of IOT research. This paper studies for this problem, and proposes analysis method of IOT end node behavior. The method collects the behavioral data of the terminal node in environmental monitoring process, analyzes data to make decisions, and then builds the sensor monitoring model with self-learning ability, and effectively ensures the security of the monitored environment. The main works are as follows:(1) A new construction method is proposed to build a sensor monitoring model to meet the special requirements of IOT monitored environment. Firstly, against the similar characteristics of the sensor network monitored environment, the o-group cluster method is used to group the nodes into many clusters by hierarchical way, which enhances the clustering effect of sample data and the accuracy of abnormal data detection. Secondly, according to the Leach algorithm the common sensor nodes of the initial cluster take turns to work, that k sensors of the cluster are chosen into working state, and other sensors into hibernation, thus effectively extending the sensor network lifetime. Finally, the base station uses Fuzzy C-means clustering algorithm to obtain the distribution of sample data by mining the collected data, and returns the results to cluster head node to detect the behavioral data of common sensor nodes, thus achieve the purpose of environmental monitoring. Simulations results show that the proposed method effectively extends the life cycle of the sensor network, and can play an effective security warning effect in assembly line production environment.(2) Against the Fuzzy C-means clustering algorithm for sensor network exists initialize data sensitive issues, a novel method is improved. Firstly, meshing method is used to collect sample data to form experimental data sets. Secondly, classification data characteristics are used, that at the center of the category, data density is relatively large and the density from the cluster center to the periphery is decreased. Methods of finding connected components are proposed to determine the initial cluster centers. Finally, the pretreated data is combined with the adaptive immune algorithm to ultimately determine the cluster centers and the number of categories. This approach makes fuzzy C-means clustering algorithm to escape from local optimal solution to get the maximum possible global optimal solution. Experimental results show that the algorithm proposed in this paper improves the convergence rate in the random distribution data sets and enhances data identification process precision and recall rate.