Security Related Data Collection And Composition In LTE Network

LTE(Long Term Evolution)is a communication technology that has been deployed in 4G networks and stepped into market due to its advantages in terms of good bandwidth,spectrum efficiency and network throughput.However,with the popularity of LTE technology,security issues in LTE networks are becoming more and more serious recently.In past ten years,researches in LTE networks are mainly focused on effective security authentication and secure access control strategies.But in today's big data era,some researches about LTE security data are published.Knowledge finding and decision processing need data as their basis for LTE network security threat and network intrusion detection.And as a key technology in data analysis,machine learning can learn rules from training data in order to classify unknown testing data.Considering the importance of LTE network security,the usage of data analysis and the effectiveness of machine learning for data analysis,in this thesis,we studied how to collect and compose LTE network security-related data by using machine learning in an efficient,correct and adaptive manner.In existing work,some literatures related to LTE network intrusion detection systems have discussed collecting LTE network security-related data in advance and using machine learning methods to identify network attacks.However,as far as we know,in a big LTE network data environment,few works discussed how to collect LTE network security-related data to avoid time cost due to repeated collection or inaccurate result due to incomplete data collection.There are also few papers discussing how to compose collected data to analyze the security issues in the whole LTE network.In order to make up the imperfection in data collection and composition,in this thesis,we propose an adaptive data collection algorithm and a composition algorithm for LTE network security measurement based on a proposed framework by us.We designed a framework for LTE network security data collection and composition.Different from traditional methods,the designed framework contains two processes: the feedback of collection strategies and the data analysis of serial-parallel structure.The two proposed algorithms are both combined with machine learning algorithms.The main algorithms that we used are feature selection algorithm and classification algorithm.Based on the two algorithms,we designed a partial mutual information gain algorithm for feature selection and a serial-parallel structure for 1-class SVM classification algorithm.The feature selection algorithm is used to compute the influence degree of LTE network security data on classification results.And then,according to the feature selection results,the data collection strategy is generated to guide future data collection.The classification algorithm is used in the serial and parallel structure to recognize and forecast which types the composed data belongs to.And then,the LTE network security measurement result will be obtained.In order to verify the performance of designed framework and algorithm,we used NS3 network simulation tools to simulate a normal LTE network environment and an abnormal LTE network environment,which contains jamming attacks,bandwidth stealing attacks and denial of service attacks.Network security data collectors are deployed in different network layers to collect related data.Then,we use Python language to implement the data collection and composition algorithms.Finally,we designed test experiments to test the performance of the proposed method.The testing results prove that the proposed algorithms have good performance when they are applied into the LTE network security analysis.