分布式协作频谱感知中针对拜占庭攻击的防御方案设计

发布时间：2018-04-25 16:53

本文选题：分布式协作频谱感知 + 拜占庭攻击　；参考：《南京邮电大学》2017年硕士论文

【摘要】：协作频谱感知是认知无线电网络中的重要技术之一。在存在阴影与衰落的感知环境中,可以极大地提高频谱感知精度。然而,协作频谱感知固有的协作性与分布特性,使其极易受到恶意用户的攻击,从而导致感知系统的不稳定。在本文中,针对分布式协作频谱感知场景中的拜占庭攻击进行了相关的防御方案设计,具体内容如下:首先,叙述了频谱感知技术的相关理论参数,介绍了在集中式与分布式场景下,协作频谱感知中常用的数据融合算法。并探讨了协作频谱感知场景中存在的拜占庭攻击类型及相应的安全防御机制。其次,研究了在分布式协作频谱感知场景中常见的几种拜占庭攻击方式,并提出了一种将效用模型引入到共识算法中的恶意攻击防御方案。在该方案中,认知用户在频谱感知过程中,被认为是智能的,有能力去计算自身的效用值,并追求自身利益最大化。通过惩罚与奖励机制,在不识别出恶意用户的情况下,使其主动放弃恶意攻击,最终使全网达成收敛一致。理论分析与仿真结果表明,本文所提方案可以有效防御多种拜占庭攻击方式。与现有的几种防御方案相比,本方案具有更好的安全性能。最后,针对认知无线电网络中恶意用户实时多变的特点,提出了一种在分布式协作频谱感知场景下,将信誉模型与强化学习结合的智能安全机制。在本方案中,每个认知用户被看成是一个智能体,通过强化学习从邻居用户中选取合作用户,并使用信誉值的形式来惩罚与奖励合作认知用户,使认知用户通过不断学习,可发现潜在的恶意用户,最终通过一致性融合算法使全网达成共识。通过数值仿真,证明了本方案可有效的甄别出恶意用户,并随着学习过程的深入,系统的安全性能会得到有效加强。
[Abstract]:Cooperative spectrum sensing is one of the most important technologies in cognitive radio networks. The spectral sensing accuracy can be greatly improved in the perceptual environment with shadow and fading. However, the inherent cooperative and distributed characteristics of cooperative spectrum sensing make it vulnerable to attack by malicious users, which leads to the instability of the sensing system. In this paper, the defense scheme of Byzantine attack in distributed cooperative spectrum sensing scene is designed. The specific contents are as follows: firstly, the related theoretical parameters of spectrum sensing technology are described. In this paper, the common data fusion algorithms in cooperative spectrum sensing in centralized and distributed scenarios are introduced. The types of Byzantine attacks in the cooperative spectrum sensing scene and the corresponding security defense mechanisms are discussed. Secondly, several common Byzantine attack methods in distributed cooperative spectrum sensing scene are studied, and a malicious attack defense scheme which introduces utility model into consensus algorithm is proposed. In this scheme, cognitive users are considered intelligent and able to calculate their own utility value and pursue the maximum of their own benefits in the process of spectrum sensing. Through the mechanism of punishment and reward, the malicious users are not identified, and the malicious attacks are given up voluntarily, so that the whole network can converge. Theoretical analysis and simulation results show that the proposed scheme can effectively defend against various Byzantine attacks. Compared with several existing defense schemes, this scheme has better security performance. Finally, in view of the real-time variability of malicious users in cognitive radio networks, an intelligent security mechanism combining reputation model with reinforcement learning is proposed in the distributed cooperative spectrum sensing scenario. In this scheme, each cognitive user is regarded as an agent. Through intensive learning, the cooperative user is selected from the neighbor user, and the credit value is used to punish and reward the cooperative cognitive user, so that the cognitive user can learn continuously. Potential malicious users can be found, and the consensus of the whole network can be reached through the consistency fusion algorithm. Through numerical simulation, it is proved that this scheme can effectively identify malicious users, and with the deepening of learning process, the security performance of the system will be effectively enhanced.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN925

【参考文献】