CCSS: Collaborative Research: Quickest Threat Detection in Adversarial Sensor Networks

近年来，随着无线通信和先进传感技术的快速发展，丰富复杂的顺序高维数据被广泛应用于入侵检测、异常检测、假新闻检测和虚假数据注入检测等威胁检测应用中。然而，对无线通信的依赖和这些联网传感器的稀疏空间分布使得它们容易受到测量操纵和虚假数据注入等对手攻击。此外，威胁通常是由人为因素引起的，因此，任何提高威胁检测算法性能的尝试都可能导致双重努力，即设计出更强大的反威胁检测技术，而留下的证据更少。在这个项目中，将开发一个博弈论框架来调查在对抗性网络环境中最快检测威胁的双重努力的最终限制。研究人员将在关于最快变化检测的会议、研讨会和研讨会上共同组织特别会议，以传播该项目的研究成果，正式确定影响深远的研究方向，确定这一新兴领域的新挑战，刺激原创研究想法的发展，并促进跨学科合作。调查人员致力于扩大未被充分代表的少数群体和妇女在研究生和本科生中对STEM教育的参与。研究人员将丰富他们现有的课程，并进一步开发与该项目相关的主题的新课程。该项目有望在最快变化检测、对抗性学习、序列分析和博弈论方面做出新的贡献。将开发一种系统的方法来开发在网络对抗环境中快速检测威胁的纳什均衡策略，并从理论上刻画它们在纳什均衡时的基本性能极限。这项工程由三个推进部分组成。第一个推力集中在具有时间结构的对抗性攻击下的一个数据流上。第二个重点是具有多个独立数据流的情况。第三个重点集中在具有图形相关结构的网络上。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

With the recent rapid development of wireless communication and advanced sensing technology, rich and complex sequential high-dimensional data are made available for a wide range of threat detection applications, e.g., intrusion detection, anomaly detection, fake news detection, and false data injection detection. However, the reliance on wireless communication and the sparsely spatial distribution of these networked sensors make them vulnerable to adversarial attacks, such as measurement manipulation and false data injection. Moreover, threats are oftentimes caused by human factors, and thus any attempt to improve the performance of threat detection algorithms may result in a dual effort to devise more powerful counter-threat-detection techniques that leave less evidence. In this project, a game-theoretic framework will be developed to investigate the ultimate limits of the dual efforts for quickest threat detection in adversarial networked environments. The investigators will co-organize special sessions at conferences, workshops, and symposia on quickest change detection to disseminate the research outcomes of this project, formalize far-reaching research directions, identify new challenges in this emerging area, stimulate the development of original research ideas, and foster interdisciplinary collaborations. The investigators are committed to broadening the participation of under-represented minorities and women both among the graduate and undergraduate students in STEM education. The investigators will enrich their current courses and further develop new courses on topics related to this project.The project is expected to make new contributions to quickest change detection, adversarial learning, sequential analysis, and game theory. A systematic methodology of developing Nash equilibrium strategies for quickest threat detection in networked adversarial environments will be developed, and their fundamental performance limits at the Nash equilibrium will be theoretically characterized. This project consists of three thrusts. The first thrust focuses on one data stream under adversarial attacks with temporal structure. The second thrust focuses on the case with multiple independent data streams. The third thrust focuses on networks with graphic correlation structure.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.