入侵检测系统在车联网中的应用研究

发布时间：2018-04-08 18:25

本文选题：车联网　切入点：入侵检测系统　出处：《深圳大学》2017年硕士论文

【摘要】：车联网(Vehicle Ad Hoc Network,VANET)被认为是能够显著地改变我们生活的新技术。通过允许网络中车辆相互通信并帮助驾驶者做出合适的决策,它能够使我们的交通行驶更加安全和舒适。由于VANET管理着关键的交通信息,并且这些交通信息和人类的生命安全密切相关,所以安全问题在VANET中至关重要。为了解决VANET中的安全问题,入侵检测系统(Intrusion Detection System,IDS)会被安装于每一辆车的内部,从而检测来自内部与外部的安全威胁。IDS通过分析来自VANET中的消息,能够检测出VANET中的异常活动,并在其它车辆出现攻击行为的时候发出警报。它是保护VANET的有效方法之一,也是近年来研究的热点。本文着重研究了IDS在VANET中的应用。首先介绍在VANET中的IDS的特点和发展;分析了业界现有的IDS算法的优势及其不足;提出新的IDS,并将其应用到VANET中。主要的贡献有:1.在VANET的IDS中,提出了新的特征抽取机制和基于GHSOM神经网络的分类器。所提出的特征提取机制用于提取两个主要特征,即交通流量特征和车辆位置特征。为了更加有效地提取位置特征,在该特征抽取机制中,本文采用了半合作方式提取位置特征。它不仅可以通过合作的方式,收集邻居车辆当前的位置信息来提取位置特征,也可以根据历史时刻的位置信息来提取位置特征。在基于GHSOM的分类器中,本文提出了额外的两步确认机制,用于对车辆消息中是否出现异常进行更准确地判断。实验表明,本文提出的IDS在其精度、稳定性、处理效率和消息规模的测量中优于其它可用IDS。2.提出一个适用于VANET下IDS的基于隐式马尔科夫模型(HMM)的滤波器模型,以减少负载和检测时间,且不损害检测的正确率。该滤波器模型预测邻居车辆的未来行为(正常或异常),以快速过滤来自这些车辆的消息。过滤模型主要包含了三个模块,分别是规划(Schedule)模块、过滤(Filter)模块和更新(Update)模块。Schedule模块根据每辆车的观察序列,采用Baum Welch算法,进行迭代地求邻居车辆的行为模式的最优HMM,并根据该HMM生成相应车辆的预测链。Filter模块根据邻居车辆的预测链,对这些车辆的未来行为进行预测和过滤。Update模块收集每次的过滤结果,然后采用时效性理论去对过滤模型进行更新,以防车辆行为模式随着时间流逝而改变。实验表明,本文提出的滤波器模型能够在不影响检测精度的条件下,使得检测时间和负载有很大的优化。
[Abstract]:Vehicle Ad Hoc Network (VANET) is considered to be a new technology that can dramatically change our lives.By allowing vehicles to communicate with each other in the network and helping drivers make appropriate decisions, it can make our traffic safer and more comfortable.Because VANET manages the key traffic information, and the traffic information is closely related to the safety of human life, the safety problem is very important in VANET.In order to solve the security problems in VANET, intrusion Detection system (IDS) is installed inside every vehicle, which detects internal and external security threats. IDS can detect abnormal activities in VANET by analyzing messages from VANET.An alarm is issued in the event of an attack by other vehicles.It is one of the effective methods to protect VANET, and it is also a hot research topic in recent years.This paper focuses on the application of IDS in VANET.This paper first introduces the characteristics and development of IDS in VANET, analyzes the advantages and disadvantages of existing IDS algorithms in the industry, and puts forward a new IDS and applies it to VANET.The main contribution is 1: 1.In IDS of VANET, a new feature extraction mechanism and a classifier based on GHSOM neural network are proposed.The proposed feature extraction mechanism is used to extract two main features, namely, traffic flow feature and vehicle location feature.In order to extract the location feature more effectively, the semi-cooperative method is used to extract the location feature in the feature extraction mechanism.It can not only collect the current location information of the neighbor vehicle but also extract the location feature according to the historical position information.In the classifier based on GHSOM, this paper proposes an additional two-step validation mechanism to determine more accurately whether there is an exception in the vehicle message.Experimental results show that the proposed IDS is superior to other available IDS.2in the measurement of its accuracy, stability, processing efficiency and message size.A filter model based on implicit Markov model (hmm) for IDS under VANET is proposed to reduce the load and detection time without compromising the accuracy of detection.The filter model predicts the future behavior of neighbor vehicles (normal or abnormal) to quickly filter messages from these vehicles.The filter model consists of three modules, namely, the planning module, filter module and update module. According to the observation sequence of each vehicle, the Baum algorithm is used.The optimal HMMs of neighbor vehicle behavior pattern are iteratively obtained, and according to the HMM, the prediction chain. Filter module of the corresponding vehicle is generated according to the prediction chain of the neighbor vehicle.The future behavior of these vehicles is predicted and the filter. Update module collects the filtering results each time. Then the filtering model is updated by time-efficiency theory in order to prevent the vehicle behavior pattern from changing with the passage of time.Experiments show that the proposed filter model can optimize the detection time and load greatly without affecting the detection accuracy.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495;TP393.08

【相似文献】