EARS: Intelligent and Cross-Layer Attack and Defense in Spectrum Sharing

随着无线应用程序和设备的激增，最近的智能手机，对无线数据的需求，也就是对无线带宽的需求，保持着指数级的增长。不幸的是，我们所有的无线视觉都需要通过公共无线电频谱来实现，这些频谱不能进一步创建，而是必须在各种无线系统和设备之间有效地共享。尽管在有效的频谱接入和共享技术方面取得了重大进展，但这些技术在实现国家宽带计划方面的成功部署，在很大程度上依赖于建立适当的安全机制。这种安全机制应该能够保护频谱共享系统中所有利益相关者的福利，特别是联邦政府和军事用户；同时，它们还应能够应对先进技术和新的接入模式带来的新挑战，这些新技术和新接入模式实现了创新的频谱共享。现阶段，频谱共享系统缺乏有效的安全机制，严重阻碍了频谱共享系统的实际部署，从而阻碍了预期的无线宽带革命的实现。该提案是填补这一空白的坚实一步，旨在更深入地了解新领域中出现的攻击策略和行为，并对频谱接入和共享的安全性提出整体看法和解决方案。预期的结果对无线行业的所有参与者以及从无线创新中受益的国民经济的所有部门都是有价值的。建议研究的跨学科性质，自然有助推动跨学科教育和全面培训未来的资讯科技人才。拟议的研究将集中在动态频谱接入（DSA）范式和认知无线电（CR）技术的安全方面，这两个方面受到了主要的研究兴趣，也提出了最重要的研究挑战。它由两个推进器组成。首先，讨论CR网络中的跨层攻击与防御。目前对CR安全的研究主要集中在单个协议层的攻击和防御，而跨层攻击自然适合CR设置，并且会造成前所未有的危害。在这个项目中，将研究一类一般的跨层攻击。针对跨层攻击，本文提出了一种跨层防御框架，重点研究了两方面的研究任务：一是推进针对CR功能独特的攻击的单独检测模块，二是指标驱动的跨层融合。在现有的研究中，假设的CR能力相当有限。随着CR技术的进一步发展，无线设备的推理和学习能力将不断增强。同时，这种技术进步也可能被对手利用，发动更复杂的恶意攻击。在第二个推动力中，假设CR用户和对手都有更多的智能和不断增长的推理和学习能力，他们之间的军备竞赛将通过随机博弈建模和多智能体强化学习（MARL）方法进行探索。该研究将通过理论分析、数值和网络模拟、原型实现以及在真实测试台上的实验来评估。

With the proliferation of wireless applications and devices, most recently the smart phones, the demand for wireless data, equivalently for wireless bandwidth, keeps an exponential growth. Unfortunately, all our wireless vision needs to be realized through the public radio spectrum, which cannot be further created but has to be efficiently shared among various wireless systems and devices. While significant progress has been made in efficient spectrum access and sharing technologies, their successful deployment towards fulfilling the national broadband plan relies crucially on the development of adequate security mechanisms. Such security mechanisms should be able to protect the welfare of all stakeholders in the spectrum sharing systems, in particular the federal government and military users; meanwhile, they should also be able to address the new challenges brought about by the advanced technologies and new access paradigms that have enabled innovative spectrum sharing. At the current stage, an effective security mechanism for spectrum sharing systems is lacking, which significantly impedes their practical deployment and consequently the realization of the anticipated wireless broadband revolution. This proposal constitutes a solid step towards filling this gap, seeking to obtain a deeper understanding of the emerging attacking strategy and behavior in this new arena, and develop a holistic view and solution to the security of spectrum access and sharing. The expected outcome will be valuable to all players in wireless industry, as well as to all sectors of the national economy that benefit from wireless innovation. The inter-disciplinary nature of the proposed research will naturally help promote cross-disciplinary education and well-rounded training of future IT workforce.The proposed research will be focused on the security aspects of the dynamic spectrum access (DSA) paradigm and cognitive radio (CR) technology, which receive major research interest and also impose foremost research challenges. It is comprised of two thrusts. First, cross-layer attack and defense in CR networks will be addressed. Current research on CR security mainly focuses on attacks and defenses at individual protocol layers, but cross-layer attack naturally fits the CR setting and can lead to unprecedented detrimental effect. In this project, a general class of cross-layer attacks will be investigated. In response to cross-layer attacks, a cross-layer defense framework is subsequently proposed, for which two research tasks will be focused on: advance in individual detection modules for attacks unique to CR functionalities, and metrics-driven cross-layer fusion. In existing studies, the assumed capabilities at CR are rather modest. With the further development of CR technology, wireless devices will grow in reasoning and learning capabilities. Meanwhile, such technology progress can also be exploited by the adversaries to launch more sophisticated malicious attacks. In the second thrust, assuming more intelligence and growing reasoning and learning capabilities for both the CR user and the adversary, the arms race between them will be explored through the stochastic game modeling and multi-agent reinforcement learning (MARL) methodologies. The proposed research will be evaluated through theoretical analysis, numerical and network simulation, prototype implementation, and experimentation on real testbeds.