The Application Of K-Means Algorithm Based On SVM In WSN Intrusion Detection System

Wireless Sensor Network(WSN)can be applied in harsh environment and special occasions,making the wireless sensor network research high practical significance.Therefore research on related problems has also been great concern in the academic circles.Different from the traditional network,the lifetime of WSN is limited and easily to be invaded,so its security problem is a relatively new topic as well as a topic to be solved.As the second protection layer of the network,the intrusion detection method of the WSN is incapable for the previous methods,so the intrusion detection research also has a great significance.On the basis of the research of various original intrusion detection technologies,this paper proposes the k-means algorithm based on SVM in WSN intrusion detection system.Firstly,this paper affirms that the application of K-means clustering algorithm in intrusion detection system is of great significance,compares and analysis of K-means algorithm and constrained-K-means algorithm with this paper algorithm.Finally,leading to this paper's algorithm to achieve better results.Above all,the mechanism initially clusters the data node set,finds the cluster centers and clusters.Then it uses support vector machine algorithm for each cluster group to maximize the distance between different types of nodes to reduce the risk of clustering,and re-divides the cluster to judge whether the cluster center is changed,finally keeps iterating until it reaches the optimal effect.The improved mechanism can better improve the detection rate and reduce the false alarm rate.This thesis takes 10% data set from Kdd cup 99 as the experimental data to carry on the experiment and then observes detection rate and false alarm rate of the improved clustering algorithm,finally makes comparison with the K-means algorithm and constrained-K-means algorithm in the rate of detection and false alarm rate.From the experimental results,it can be seen that whatever a set of training in a single type of known attack or several types of unknown mixed data sets,compared with the K-means algorithm and constrained-K-means algorithm,this mechanism has higher detection rate and lower false alarm rate.It performs better in the intrusion detection.And it's a completely unsupervised lightweight algorithm for WSN.