Research On Intrusion Detection Method For Wireless Sensor Networks Based On Traffic

With the rapid development of big data,cloud computing and Internet of Things technologies,Wireless Sensor Network(WSN)has been deeply researched and widely used as a new type of network.Because the sensor nodes in WSN have limited computing,storage and energy,they are often deployed in complex network environments where personnel cannot reach and conditions are harsh.Therefore,WSN faces various challenges in terms of security and must use an effective protection mechanisms to ensure WSN security.Network traffic anomalies are characterized by unknown parameter characteristics and sudden traffic generation.Some typical attack methods in WSN,such as Wormhole,Sinkholes,Hello Flooding,and Jamming,cause network traffic to deviate from normal traffic,which will bring huge harm to the WSN system in a short time.Therefore,it is particularly important to accurately and quickly detect network traffic anomalies and establish a suitable traffic-based WSN intrusion detection system model.Aiming at the above problems,the thesis mainly uses the information gain ratio feature selection method,ensemble learning,random forest and deep forest and other machine learning methods to study the key technologies of WSN intrusion detection system.The main research work is as follows:1.Aiming at the problem that the dimension of the traffic data to be processed in the WSN intrusion detection method is too high,which leads to the large amounts of computational complexity of the intrusion detection model and the weak detection performance of the intrusion behavior.An intrusion detection model based on information gain ratio and Bagging algorithm was proposed by using the principle of ensemble learning algorithm.Firstly,the information gain ratio method is used to select the feature of sensor node traffic data in this model.Secondly,the Bagging algorithm is used to construct an ensemble classifier so as to train multiple C4.5decision trees.The parameters of the ensemble classifier are optimized through 10 iterations,and the dynamic pruning process is introduced.Finally,the classification results of C4.5 decision tree are classified and detected by majority voting mechanism.The experimental results show that compared with the existing intrusion detection methods,the proposed model has high detection accuracy for the intrusion attack behavior.While ensuring the detection rate of 99.4%,it can still maintain a low false alarm rate and high detection performance for intrusion behavior.2.Aiming at the problems of poor performance,poor real-time detection and high complexity of the existing feature selection algorithm and classification algorithm in WSN intrusion detection system,a distributed WSN intrusion detection model based on random forest and deep forest algorithm is proposed.Firstly,the sensor node traffic data are preprocessed by this model,then the lightweight random forest classifier is deployed to the sensor node and cluster head node.The sensor node and the cluster head node cooperate to process the traffic data.The deep forest algorithm is used on the base station to find the attack behavior from a large amount of traffic data.Finally,it performs real-time classification intrusion detection on the intrusion behavior in WSN.The experimental results show that compared with the existing intrusion detection model,the proposed model has good detection performance and high real-time performance,which can avoid over-fitting of the model.