CRII: CNS: Towards Robust and Efficient Dynamic Spectrum Sharing with Knowledge Transfer

无线设备的激增和对无线服务日益增长的需求，再加上低效的频谱分配和有限的频谱可用性，加剧了下一代网络(NextG)中无线频谱资源的稀缺。该项目专注于开发用于动态频谱共享(DSS)的新型转移学习(TL)框架，以实现跨用户、环境和无线系统的知识转移，为更有效地智能地利用未充分利用的许可频谱提供可行的方法。DSS无线系统具有动态环境、异质网络、海量连接、干扰、高通信开销、有限的计算和存储能力以及安全和隐私问题等特点，这使得学习和利用可转移知识具有挑战性。此外，要达到预期的知识转移效果，往往需要大量高质量的培训数据，而转移数据知识可能会引起安全和隐私问题，限制了对其他任务的适应和推广。因此，本项目旨在探索新的目标语言学习策略，以学习可转移的知识，并解决与DSS系统中的稳健性、效率、安全性和隐私有关的问题。该项目的一个主要目的是对目标DSS无线系统的特征和参数进行系统调查，同时探索与知识转移相关的基本原则、理论和独特的挑战。这些研究旨在弥合系统特性和算法开发之间的差距。(1)设计了一个集成评估方案来评估基于目标语言的决策支持系统框架的健壮性、效率、安全性和保密性。(2)开发高效的基于TL的DSS框架，用于自适应频谱感知、选择、接入和切换。(3)创建强大的安全和隐私TL策略，用于监控、检测、缓解和防止各种恶意攻击，同时保护敏感数据。与此同时，研究团队正在开发一种无线知识转移试验台，其中包括可转移的知识、评估方案、预先训练的TL模型、攻击知识库以及安全和隐私策略。该实验平台有助于促进和规范无线通信系统中知识重用的研究。研究和教育计划的整合为DSS、人工智能、迁移学习和网络安全领域的NextG劳动力做好了准备。推广活动通过不同的学习方式在DSS研究和K-12学生、少数群体和大学生之间建立联系。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。

The proliferation of wireless devices and the increasing demand for wireless services, coupled with inefficient spectrum allocation and limited spectrum availability, exacerbate the scarcity of wireless spectrum resources in next-generation (NextG) networks. This project focuses on the development of novel transfer learning (TL) frameworks for dynamic spectrum sharing (DSS) to enable knowledge transfer across users, environments, and wireless systems, offering viable approaches to intelligently utilize underutilized licensed spectrum more effectively. DSS wireless systems exhibit characteristics such as dynamic environments, heterogeneous networks, massive connections, interference, high communication overhead, limited computing and storage capacity, as well as security and privacy concerns, making it challenging to learn and leverage transferable knowledge. Moreover, achieving the desired performance of knowledge transfer often requires substantial amounts of high-quality training data, while transferring data knowledge may raise security and privacy issues, limiting adaptation and generalization to other tasks. Therefore, this project aims to explore novel TL strategies for learning transferable knowledge and addressing concerns related to robustness, efficiency, security, and privacy in DSS systems. A key thrust of the project involves a systematic investigation into the characteristics and parameters of target DSS wireless systems, alongside an exploration of the fundamental principles, theories, and unique challenges associated with knowledge transfer. These studies aim to bridge the gap between system characteristics and algorithm development. The research tasks include the following: (1) Design an ensemble evaluation scheme to assess the robustness, efficiency, security, and privacy of TL-based DSS frameworks. (2) Develop efficient TL-based DSS frameworks for adaptive spectrum sensing, selection, access, and handoff. (3) Create robust security and privacy TL strategies for monitoring, detecting, mitigating, and preventing various malicious attacks, while also protecting sensitive data. Concurrently, the research team is developing a Wireless Knowledge Transfer testbed that incorporates transferable knowledge, evaluation schemes, pre-trained TL models, attack knowledge databases, and security and privacy strategies. This testbed helps to facilitate and standardize research on knowledge reuse in wireless communication systems. The integration of research and education plans prepares the NextG workforce in the fields of DSS, artificial intelligence, transfer learning, and cybersecurity. Outreach activities establish connections between the DSS research, and K-12 students, minority groups, and college students through various learning approaches.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.