Research On AI-based Intrusion Detection In Software Defined Networks

In recent years,the extensive and deep application of information technology has led to an increasingly prominent network security problems.The rise of Software Defined Networking(SDN)has promoted the development of network information security.Most existing intrusion detection systems and devices are deployed independently and are difficult to coordinate with each other.In addition,they are based on feature matching and pattern comparison,which is difficult to intelligently identify unknown attacks.The introduction of Artificial Intelligence(AI)technology overcomes the drawbacks and limitations of traditional IDS.Based on Artificial Intelligence algorithms,this paper studies the architecture and technology of network intrusion detection in the environment of Software Defined Networks.In this paper,we first introduce the concept and model of intrusion detection and further elaborate the categories and common techniques of intrusion detection.Then,we explain the data set and evaluation metrics used in the experiments.Taking the advantages of Software Defined technology in network security,this paper proposes an intelligent intrusion detection system based on machine learning algorithms.In the intelligent layer of the system,it applies Random Forest(RF)to select typical features,and use the combination of k-means++ and Adaboost to perform flow-based classification to detect network intrusions.Compared with existing systems,we verify the effectiveness and optimization of the intrusion detection system.The paper also proposes a two-stage intrusion detection technology based on improved AI algorithms.In the two critical steps of feature selection and flow classification,we enhance Bat Algorithm(BA)and Random Forest respectively.The optimized algorithms and their combination are trained and simulated based on the public data set and compared with their original algorithms as well as existing machine learning algorithm combinations:The improved algorithms can select optimal feature subsets more effectively and it enhances the ability of flow classification,improving the detection accuracy of various types of attacks.