---

Project Description

Wireless network security designs have been traditionally guided by theoretical or decision logic frameworks, such as estimation and detection theory, protocol control mechanisms , and empirical analysis to discover vulnerabilities. Today's wireless network operations are facing increasingly more complicated and highly-dense environments, which generate diverse operational data in large volumes (e.g., interference measurements, protocol control data, wireless channel state information). As a result, significant academic efforts have been recently focused on adopting artificial intelligence (AI) techniques for 5G and NextG security because of their efficiency and capability of handling various data to achieve intelligent functionalities. Despite the wide adoption of AI for wireless network security solutions, the educational efforts are not in pace with its research counterpart. In particular, we review the related work in the educational context of wireless network security and AI in the public domain.

This proposed project, by University of South Florida (USF) and University of Oklahoma (OU), focuses on building a comprehensive educational program of AI for 5G/NextG wireless network security with lab-based curriculum modules and project-based training modules. Lab-based curriculum modules consist of the physical layer module, the medium access control (MAC) and network layers module, and the network applications module, which are classified according to the standard wireless protocol stack. We aim to achieve balanced wireless network content throughout the protocol stack. Each individual module is focused on demonstrating a particular wireless topic of using AI techniques in four categories of classical machine/deep learning, decentralized learning, dynamic learning and recurrent learning. Moreover, the project-based training is aimed to provide students with a semester-long training experience that involves more in-depth design and evaluation of using AI for wireless network security, as compared to the lab-based modules that offer introductory knowledge during lectures. The projects are designed to provide students with several potential exploration directions to apply their learned knowledge towards solving relatively complex wireless network security problems using AI techniques.

---

Personnel

• Principle Investigators:

  • Zhuo Lu and Yao Liu at University of South Florida
  • Shangqing Zhao and Hazem Refai at University of Oklahoma

• Students:

  • Wenwei Zhao
  • Jinmiao Chen

---

Results and Products

• An AI perspective for Attacks: download
• Leveraging RF Leakage for Eavesdropping Detection by AI: download
• Model Poisoning Attacks against Federated Learning in IoT: download

---

Broader Impacts

• The PIs have integrated the initial training materials in their classes: Cryptography and Data Security and Wireless Mobile Computing and Security at USF, and Wireless and Mobile Networks at OU. These modules help the students better understand the emerging topics of using machine learning in wireless network security designs.

• The PIs hold regular discussions with the graduate students in this project. Students who previously focused on one or few areas of their research topics also gain gradually the knowledge outside their primary area related to using AI and machine learning techniques in wireless network scenarios from the physical layer to the application layer.

• The project involves two female Ph.D. students.