Machine Learning for Intrusion Detection into Unmanned Aerial System 6G Networks

Author

Faisal Alrefaei, Embry-Riddle Aeronautical UniversityFollow

Date of Award

Spring 5-7-2024

Access Type

Dissertation - Open Access

Degree Name

Doctor of Philosophy in Electrical Engineering & Computer Science

Department

Electrical Engineering and Computer Science

Committee Chair

Radu F. Babiceanu

Committee Co-Chair

Houbing Song

First Committee Member

Laxima Niure Kandel

Second Committee Member

Eduardo A. Rojas-Nastrucci

Third Committee Member

Kenji Yoshigoe

Abstract

Progress in the development of wireless network technology has played a crucial role in the evolution of societies and provided remarkable services over the past decades. It remotely offers the ability to execute critical missions and effective services that meet the user's needs. This advanced technology integrates cyber and physical layers to form cyber-physical systems (CPS), such as the Unmanned Aerial System (UAS), which consists of an Unmanned Aerial Vehicle (UAV), ground network infrastructure, communication link, etc. Furthermore, it plays a crucial role in connecting objects to create and develop the Internet of Things (IoT) technology. Therefore, the emergence of the CPS and IoT technologies provided many connected devices, generating an enormous amount of data. Consequently, the innovation of 6G technology is an urgent issue in the coming years. The 6G network architecture is an integration of the satellite network, aerial networks, terrestrial networks, and marine networks. These integrated network layers will provide new enabling technologies, for example, air interfaces and transmission technology. Therefore, integrating heterogeneous network layers guarantees an expansion strategy in the capacity that leads to low latency, ultra-high throughput, and high data rates. In the 6G network, Unmanned Aerial Vehicles (UAVs) are expected to densely occupy aerial spaces as UAV flying base stations (UAV-FBS) that comprise the aerial network layer to offer ubiquitous connectivity and enhance the terrestrial network in remote areas where it is challenging to deploy traditional infrastructure, for example, mountain, ocean deserts, and forest. Although the aerial network layer offers benefits to facilitate governmental and commercial missions, adversaries exploit network vulnerabilities to block intercommunication among nodes by jamming attacks and violating integrity through executing spoofing attacks. This work offers a practical IDS onboard UAV intrusion detection system to detect unintentional interference, intentional interference jamming, and spoofing attacks. Integrating time series data with machine learning models is the main part of the suggested IDF to detect anomalies accurately. This integration will improve the accuracy and effectiveness of the model. The 6G network is expected to handle a high volume of data where non-malicious interference and congestion in the channel are similar to a jamming attack. Therefore, an efficient anomaly detection technique must distinguish behaviors in the drone's wireless network as normal or abnormal behavior. Our suggested model comprises two layers. The first layer has the algorithm to detect the anomaly during transmission. Then it will send the initial decision to the second layer in the model, including two separated algorithms, confirming the initial decision separately (nonintentional interference such as congestion in the channel, intentional interference jamming attack, and classify the type of jamming attack, and the second algorithm confirms spoofing attack. A jamming attack is a stealthy attack that aims to exhaust battery level or block communication to make wireless UAV networks unavailable. Therefore, the UAV forcibly relies on GPS signals. In this case, the adversary triggers a spoofing attack by manipulating the Global Navigation Satellite System (GNSS) signal and sending a fake signal to make UAVs estimate incorrect positions and deviate from their planning path to malicious zones. Hackers can start their malicious action either from malicious UAV nodes or the terrestrial malicious node; therefore, this work will enhance security and pave the way to start thinking about leveraging the benefit of the 6G network to design robust detection techniques for detecting multiple attacks that happen separately or simultaneously.

Scholarly Commons Citation

Alrefaei, Faisal, "Machine Learning for Intrusion Detection into Unmanned Aerial System 6G Networks" (2024). Doctoral Dissertations and Master's Theses. 815.
https://commons.erau.edu/edt/815