Public defence in Communications Engineering and Networking Technology, M.Sc. Yongchao Dang

The title of the thesis: Machine Learning based GNSS Spoofing Detection and Mitigation for Cellular-Connected UAVs

Doctoral student: Yongchao Dang
Opponent: Prof. Petri Välisuo, University of Vaasa, Finland
Custos: Prof. Riku Jäntti, Aalto University School of Electrical Engineering, Department of Information and Communications Engineering 

Cellular-connected Unmanned Aerial Vehicle (UAV) systems are envisioned as the composed part of the upcoming 5G-and-beyond mobile cellular networks by providing delivery applications and emergency communications. With the help of the mobile cellular network and Global Navigation Satellite System (GNSS), cellular-connected UAVs can be easily deployed to remote or dense areas on demand. However, the civil GNSS service is unencrypted and vulnerable to spoofing attacks, which threatens the security of cellular-controlled UAVs. Therefore, the 3rd Generation Partnership Project (3GPP) has initiated a set of techniques and supports that enable mobile cellular networks to track and identify UAVs to enhance low-altitude airspace security. 

In this thesis, we explore the utilization of machine learning methodologies and 3GPP strategies to effectively identify and counteract instances of GNSS spoofing, with a specific focus on GPS manipulation. The thesis investigates the designs of deep neural networks and graphic neural networks to detect GPS spoofing for cellular-connected UAVs or UAV swarms, and the integration of 3D radio mapping and particle filtering to recover UAV positions and mitigate the GPS spoofing. 

The proposed neural networks could be deployed within the 5G-assisted Unmanned Aerial System (UAS) framework to detect and mitigate GPS spoofing attacks. The experimental results suggest that Convolutional Neural Networks (CNNs) are recommended for spoofing detection because they can uncover key spoofing patterns efficiently by processing the temporal and spatial structure of data. Besides, the Graphic Neural Network (GNN) can monitor UAV swarms and detect GPS spoofing within one second. Moreover, when combined with 3D radio mapping and particle filtering, it enables accurate recovery of UAV positions after spoofing incidents. Thus, the proposed machine learning methods can detect and mitigate GNSS spoofing attacks for cellular-connected UAVs, thereby ensuring the reliability and safety of cellular-connected UAV operations.

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact information:

See also:

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53