随着无线通信技术不断快速发展，无线频谱资源匮乏的问题愈发受到重视。认知无线电频谱感知技术能够帮助提高频谱利用率，改善效率低下的传统频谱分配方式。多用户协作频谱感知技术的提出，帮助认知无线电网络克服衰落环境，改善频谱感知性能。为保证协作频谱感知系统的安全性，需要研究协作频谱感知抗恶意节点算法。首先，本文介绍了认知无线电技术，它是智能地实现授权频段内非授权用户伺机接入频段的技术。在认知无线电技术中，首先要解决是感知主用户是否存在的问题。本文介绍了频谱感知技术的性能参数，对本文使用的频谱感知方法——能量检测进行算法研究。搭建频谱感知模型，仿真分析了单用户频谱感知在衰落环境下的缺陷，以及多用户协作频谱感知在衰落环境影响的性能优势。其次，本文分析了协作频谱感知系统中可能存在的安全问题。针对恶意节点对数据融合过程的攻击，引入拜占庭将军问题进行阐述。拜占庭攻击是个入侵容忍体系问题，它要求在存在恶意节点的情况下，让其他诚实节点做出正确的全局决策。在该体系中，分析了可能存在的恶意节点类型和潜在的攻击方式。通过仿真研究恶意节点的攻击对协作频谱感知系统造成的不良影响。仿真结果显示，协作频谱感知的性能随着恶意节点所占比例的增大而降低，造成感知效率低下。然后，为解决协作频谱感知系统的拜占庭攻击问题，本文研究了抗恶意节点攻击的算法。抗恶意节点算法需要实现对恶意节点的检测，同时消除恶意节点的恶意影响。采用了基于异常值检测的算法和基于构造权重函数算法，研究算法原理，进行仿真分析。基于异常值检测的算法是结合数据统计中的异常值检测方法来实现去除恶意节点的影响；基于构造权重函数的算法是通过构造可信参数函数和信誉度函数来实现对所有认知节点进行权重评估，并且在融合中心对所有节点的感知结果进行数据融合。通过仿真研究发现，两种算法都能实现一定程度的抗恶意节点效果，算法提升了协作频谱感知系统抵抗恶意节点的能力。最后，通过对算法的分析，提出抗恶意节点的算法改进方案——在基于构造权重函数的算法中提出对本地感知数据的预处理和在基于异常值检测的算法中提出增大数据控制信道的量化比特。仿真结果表明，算法改进措施能够实现性能提升。本文的研究表明，虽然恶意节点的拜占庭攻击会对协作频谱感知系统造成恶劣影响，但是本文中的抗恶意节点的算法和改进措施能够有效消减恶意节点的影响，提升协作频谱感知系统的性能，从某种程度上也增强了认知无线电网络的安全性。

With the continuous development of wireless communication, spectrum resource scarcity is getting more and more attention. Cognitive radio can alleviate the inefficient frequency assignment problem and improve the utilization of licensed spectrum bands.Cooperative spectrum sensing for cognitive radio system outperforms single user spectrum sensing in many aspects. However, cooperative spectrum sensing may be attacked by malicious users. In this paper, methods to against malicious attack in cooperative spectrum sensing system are studied. Firstly，the thesis introduces cognitive radio system, in which unlicenced users sense the licenced user and then opportunistically dynamically access the licenced spectrum band when it is not occupied. Spectrum sensing technique which identifies the idle bands is a core issue in cognitive radio system. Performance parameters for spectrum sensing and energy detection technique are presented in detail in the thesis. A model of spectrum sensing in deep fading enviroment is built to analysis disadvanages of single-user sensing and advantages of multiple-user cooperative sensing. Secondly, potential secure problems in cooperative sensing system are studied in the paper. The problem malicious nodes attacking fusion center is considered as a Byzantine attack problem, in which we need to eliminate the effects of malicious nodes and let the fusion center give the proper results using the information from honest nodes. Simulation results in this paper show that potential malicious nodes with different attack modes may significantly decrease the performance of cooperative sensing system. Results show that the higher the malicious nodes ratio is, the worse the sensing performance is. And then, methods which are used to identify malicious nodes and to eliminate the effects of malicious nodes are studied. Methods based on outlier detection and weighting function are applied to solve Byzantine attack problem, which are studied in detail in this paper. We can see the positive effects from simulation results by using methods upon. Methods against Byzantine attacks improve the sensing performance of cooperative sensing system with Byzantine attack. Finally, two method improvements are introduced: one is that each secondary user takes a pre-progress of local detection data; the other one is choosing a bigger quantization bits for control channel. By simulating the cooperative sensing system with Byzantine attack，we can see that method improvements can make further improvement on system performance. Simulation results show malicious nodes’ Byzantine attack will decrease performance of cooperative sensing system and methods against malicious attack present in this paper can improve the performance to a certain degree. Analysis indicates that methods in this paper are robust against malicious attacks and can successfully raise cognitive radio system security.