Home // VEHICULAR ANALYTICS 2024, The First International Conference on Vehicular Systems // View article

Unmasking Threats in UAV Networks: A Semi-Supervised Approach to Cyphal Security

Authors:
Kabid Hassan Shibly
Ryoichi Isawa
Takahiro Kasama

Keywords: unmanned aerial vehicle; cyphal; in-vehicle networks; cybersecurity; intrusion detection systems; semi-supervised learning

Abstract:
With increased connectivity in In-Vehicle Networks (IVNs), protocols like Cyphal, used in Unmanned Aerial Vehicles (UAVs), are vulnerable to cyber threats, including flooding, fuzzy, and replay attacks. Traditional Intrusion Detection Systems (IDS) rely on supervised learning and struggle with evolving attacks due to the need for large volumes of labeled data. We propose Gravity Well Learning (GWL), a novel semi-supervised learning framework for intrusion detection in Cyphal networks. GWL leverages both labeled and unlabeled data to enhance detection accuracy while reducing reliance on extensive labeled datasets. It introduces a central "Planet" model, guided by expert "Gravity Wells" that refine detection capabilities. Experiments show that GWL achieves 65.50% accuracy with 10% labeled data and 83.10% with 40%. These results underscore GWL's robustness and scalability in securing UAV and automotive networks, making GWL a promising solution for real-world intrusion detection in vehicular communication systems.

Pages: 1 to 8

Copyright: Copyright (c) IARIA, 2024

Publication date: November 17, 2024

Published in: conference

ISBN: 978-1-68558-217-3

Location: Valencia, Spain

Dates: from November 17, 2024 to November 21, 2024