Research On Key Technology Of LTE Intrusion Detection Systems

With traditional mobile network unable to meet the growing data capacity, LTE (Long Term Evolution) technology as the latest mobile technology standards become more sophisticated, and will soon be widely used. But the network security issues it faces become increasingly prominent. Traditional encryption and firewall technology has been unable to fully meet the security needs at the same time, so intrusion detection technology are becoming more and more important as a new security measure. The intrusion detection system can recognize and response the malicious use of computer and network. It is a security technology which protects the network from hackers and a new generation of network security technology after firewall, data encryption, and other traditional security measures. The new attack methods will continue to appear in the new network environment which brings new issues to mobile network intrusion detection research.A simple cell phone firewall has been unable to prevent this type of invasion effectively, so in this paper we set the intrusion detection system in the eNodeB (base station nodes) platform, in the LTE network, we can effectively detect intrusions in eNodeB status and network traffic monitoring.In this paper, we design and establish the key part of intrusion detection system based on data mining technology. This paper introduces the key technologies and system architecture of LTE, then elaborated on the concept of intrusion detection, related technologies, and research status. We analyze the application of data mining technology in the Intrusion Detection System. On this basis, we design and implement a data mining-based intrusion detection system. The system is mainly divided into two parts:Flow detection based on single eNodeB.Using wireshare plugin interface library ,We develop a progame to capture all the data packets meeting the 3GPP protocol , monitor network eNodeB behavior, and write specific network packet to the log.This paper presents text mining algorithms using parts of speech as a reference , which can effectively discover the association rules among strings.Based on Bootstrapping algorithm, it not only reduce the dependence on stemming in pre-processing stage ,but also can handle the Chinese vocabulary appeared in the log . It increase the understanding of content of the log, improve the effectiveness of association rules, effectively improve mining efficiency and provide association rules in application of log mining.Eigenvalues detection based on multi-eNodeB. Using unique Traces system of eNodeB ,we obtain all UDP packets which record all features of the current eNodeB. Using protocols identification and double-buffer flow of reading and writing techniques, this paper presents a improved LE based on non-linear popular study.It defines the different measure way in class and between classes, replaces the Euclidean distance measure between the sample points in the original algorithm, retains the high efficiency of Labourasse feature maps and effective reduces the classification costs of multi-characteristic value. Then it use characteristic value after dimension reduction as input of BP neural network, which reduces the training time of neural networks and discovers the association between eigenvalue and intrusion effectively.