基于DBN的移动自组织网络入侵检测技术研究

发布时间：2018-11-08 18:01

【摘要】：移动自组织网络在无线通信领域得到了越来越广泛的应用,但其固有特性使其容易受到各种各样的入侵,对其安全性研究是非常有价值的。入侵检测技术作为一种主动式安全防护机制,是保障移动自组织网络安全的关键。深度学习主要探讨多层人工神经网络的建模和学习问题,已经在语音、图像识别等领域获得了巨大成功,为解决移动自组织网络入侵检测这样一个复杂的行为识别问题提供了新的有效途径。针对移动自组织网络安全问题的多样性和复杂性,本文提出了一种基于深度信念网络(Deep Belief Network,DBN)的入侵检测方法。DBN是一种比较成熟的深度学习模型,将DBN应用于移动自组织网络入侵检测技术中,能够取得较高的检测正确率。本文的主要工作和贡献在于:结合移动自组织网络的特点和安全威胁,本文首先分析了入侵检测技术应用于移动自组织网络面临的难题、几种典型的入侵检测算法和模型。研究了DBN模型的学习原理,受限玻尔兹曼机的训练算法;分析了将DBN应用于移动自组织网络入侵检测技术中原理上的可行性。其次,本文设计了基于DBN的移动自组织网络入侵检测模型的体系结构,包括对无线数据包捕获,数据预处理,模型训练和入侵检测等模块的详细设计;给出了DBN模型训练过程中会遇到的几个问题的解决方法。最后,对本文提出的基于DBN的移动自组织网络入侵检测方法进行了仿真。针对移动自组织网络路由层的拒绝服务攻击,本文在NS2中添加了黑洞节点和自私节点,模拟两种网络入侵。分别对正常网络和加入攻击节点的网络进行了仿真,分析了网络性能,并提取了网络行为特征。基于MATLAB仿真了DBN入侵检测模型,并对其进行训练和测试。测试结果验证了本文提出的基于DBN的移动自组织网络入侵检测方法的可行性;与传统BP神经网络入侵检测方法进行比较,DBN具有更好的入侵检测性能。
[Abstract]:Mobile ad hoc networks have been more and more widely used in the field of wireless communication, but their inherent characteristics make them vulnerable to various kinds of intrusion, so it is very valuable to study their security. As an active security protection mechanism, intrusion detection technology is the key to ensure the security of mobile ad hoc networks. Depth learning mainly discusses the modeling and learning problems of multi-layer artificial neural networks, and has been a great success in speech, image recognition and other fields. It provides a new and effective way to solve the complex behavior recognition problem of mobile ad hoc network intrusion detection. In view of the diversity and complexity of security problems in mobile ad hoc networks, this paper proposes an intrusion detection method based on deep belief network (Deep Belief Network,DBN). DBN is a mature deep learning model. The application of DBN in mobile ad hoc network intrusion detection technology can achieve a high detection accuracy. The main work and contributions of this paper are as follows: considering the characteristics and security threats of mobile ad hoc networks, this paper first analyzes the problems faced by the application of intrusion detection technology in mobile ad hoc networks, and several typical intrusion detection algorithms and models. The learning principle of DBN model and the training algorithm of constrained Boltzmann machine are studied, and the feasibility of applying DBN to mobile ad hoc network intrusion detection technology is analyzed. Secondly, this paper designs the architecture of mobile ad hoc network intrusion detection model based on DBN, including the detailed design of wireless packet capture, data preprocessing, model training and intrusion detection module. Some solutions to the problems encountered in DBN model training are given. Finally, the proposed intrusion detection method based on DBN in mobile ad hoc networks is simulated. In view of the denial of service attack in the routing layer of mobile ad hoc networks, black hole nodes and selfish nodes are added to NS2 to simulate the two kinds of network intrusion. The normal network and the network with attack nodes are simulated, the network performance is analyzed, and the network behavior characteristics are extracted. The DBN intrusion detection model is simulated, trained and tested based on MATLAB. The test results verify the feasibility of the proposed mobile ad hoc network intrusion detection method based on DBN, and compared with the traditional BP neural network intrusion detection method, DBN has better intrusion detection performance.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN915.08

【共引文献】