基于滑动窗口的协作频谱感知对抗拜占庭攻击

Sliding window-based cooperative spectrum sensing against Byzantine attack

摘要:

认知无线电技术允许从用户动态地接入主要用户被授权的频谱，提高频谱利用率。协作频谱感知是认知无线电技术的一个重要组成部分，通过空间分集检测主用户信号。然而，由于认知无线网络的开放性，协作频谱感知过程可能会受到拜占庭攻击，恶意用户伪造有关主用户信号的状态信息，然后对主用户的通信造成干扰或自私地占用频谱资源，此外，协作频谱感知因多个从用户协作而需要更多的时间来检测主用户信号，因而将导致协作频谱感知的性能和效率进一步降低。针对上述问题，提出了基于滑动窗口的协作频谱感知方案，以减轻拜占庭攻击的负面影响，提高协作效率。在深入分析融合中心盲的问题的基础上，从恶意用户的角度出发，建立了一个随机拜占庭攻击模型来描述恶意行为。为了解决感知样本融合过程中的盲的问题，提出了一种交付评估机制，为基于滑动窗口的协作频谱感知奠定了坚实的基础，并在一个滑动窗口内进一步评估信誉值，以提高报告阶段的协作效率。仿真结果表明，无论恶意比例如何，基于滑动窗口的协作频谱感知在始终攻击的情况下只需要6个平均样本数就可以提供100%的检测准确率，而在恶意比例超过50%的随机攻击的情况下依然能够展现出显著的性能优势。

Abstract:

Cognitive radio (CR) allows secondary users (SUs) to dynamically access the spectrum resources being authorized by the primary user (PU) and improves the spectrum utilization. Cooperative spectrum sensing (CSS) is the key function of CR technology to detect the PU signal by spatial diversity. However, due to the openness of CR networks (CRNs), the CSS process may suffer from Byzantine attack, in which malicious users (MUs) falsify the state information about the PU and then cause the harmful interference to the PU’s communication or selfishly occupy the spectrum resources. In addition, CSS requires more sensing times to detect the PU signal because of the cooperative paradigm, therefore further decreasing the cooperative performance and efficiency. In view of this, this paper proposed a sliding window-based CSS (SW-CSS) scheme to mitigate the negative impact of Byzantine attack and improve cooperative efficiency. To this end, on basis of in-depth analyses on the blind problem, this paper formulated a random Byzantine attack model from the malicious perspective to characterize the malicious behaviors. In order to solve the blind problem in the process of the sample fusion, this paper proposed a delivery evaluation mechanism to lay a solid foundation for SW-CSS. On the basis of this, this paper further evaluated the trust value (TrV) to improve the cooperative efficiency in the reporting stage. At last, simulation results show that regardless of the malicious ratio, SW-CSS only requires the average number of samples (ANS) 6 to provide with 100% detection accuracy in the presence of always attack while also provides with remarkable performance when the malicious ratio exceeds 50%.