A Federated Learning-based Lightweight Network with Zero Trust for UAV Authentication

Hao Zhang, Fuhui Zhou, Wei Wang, Qihui Wu, Chau Yuen · 2025

Read A Federated Learning-based Lightweight Network with Zero Trust for UAV Authentication by Hao Zhang, Fuhui Zhou, Wei Wang, Qihui Wu, Chau Yuen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-05111.txt.

Introduction

Hao Zhang and four co-authors present a 2025 system for checking the identity of drones in a network while keeping data private. Drones move fast and face jamming, spying, and cyber attack, so a network must confirm which drones are genuine and reject fakes. Their system learns from radio spectrograms, pictures of a signal's frequency pattern, using a small neural network. Training is federated, meaning the drones share model updates rather than raw data, and access follows a zero-trust rule.

In tests, the network scored above 80 percent accuracy on drone types it knew. It also flagged unknown drone types fairly well. The model stayed small, with low storage needs, compared with older benchmarks. Tests varied the number of training clients and the share of unknown drones, and the system held up well.

Catalog record

CREATORHao Zhang, Fuhui Zhou, Wei Wang, Qihui Wu, Chau Yuen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-05111

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Federated Learning-based Lightweight Network with Zero Trust for UAV Authentication Hao Zhang, Fuhui Zhou, Wei Wang, Qihui Wu, Chau Yuen Abstract Unmanned aerial vehicles (UAVs) are increasingly being integrated into next-generation networks to enhance communication coverage and network capacity. However, the dynamic and mobile nature of UAVs poses significant security challenges, including jamming, eavesdropping, and cyber-attacks. To address these security challenges, this paper proposes a federated learning-based lightweight network with zero trust for enhancing the security of UAV networks. A novel lightweight spectrogram network is proposed for UAV authentication and rejection, which can effectively authenticate and reject UAVs based on spectrograms. Experiments highlight LSNet’s superior performance in identifying both known and unknown UAV classes, demonstrating significant improvements over existing benchmarks in terms of accuracy, model compactness, and storage requirements. Notably, LSNet achieves an accuracy of over 80 % 80\% for known UAV types and an Area Under the Receiver Operating Characteristic (AUROC) of 0.7 0.7 for unknown types when trained with all five clients. Further analyses explore the impact of varying the number of clients and the presence of unknown UAVs, reinforcing the practical applicability and effectiveness of our proposed framework in real-world FL scenarios. I Introduction R ECENT developments in wireless communication technology have significantly expanded the use and capabilities of wireless devices, leading to an increased demand for such services. Unmanned aerial vehicles (UAVs) are increasingly becoming crucial components of next-generation networks, enhancing communication coverage and boosting the capacity of terrestrial base stations [ 1 ] . Additionally, UAVs are anticipated to play a pivotal role in the architecture of the sixth-generation (6G) networks, particularly in densely populated areas. These UAVs enable the rapid and efficient scalability of the 6G network to accommodate fluctuating demands. They can also provide essential wireless backhaul for small cells or support the broader 6G network infrastructure [ 2 ] . Moreover, UAVs hold promise in fostering the development of more dynamic, engaging, and interactive virtual environments, crucial for powering metaverse applications and services by leveraging their aerial capabilities and processing power. However, the integration of UAVs into 6G networks poses significant security challenges. UAVs are vulnerable to various security threats, including jamming, eavesdropping, and cyber-attacks. These threats can compromise the confidentiality, integrity, and availability of the data transmitted by UAVs, thereby undermining the overall security of the network. To address these security challenges, it is essential to develop robust security mechanisms that can protect UAVs from potential security threats. As shown in Fig. 1 , UAV authentication methods can be classified into two categories, namely, traditional authentication mechanisms and non-contact UAV detection techniques. Traditional authentication mechanisms, also known as access control mechanisms, identify and verify the identity of UAVs and ground stations by using physical unclonable functions (PUFs) [ 3 , 4 ] , traffic data analysis [ 5 , 6 ] , and other methods. However, traditional access control mechanisms are not well-suited for UAV networks due to their dynamic and mobile nature. Moreover, it is hard to design a lightweight and robust PUF-based authentication mechanism that can adapt to the dynamic and mobile nature of UAV networks. Apart from these traditional authentication mechanisms, non-contact UAV detection techniques have been proposed to detect and classify UAVs based on their unique features. Physical signals include acoustic [ 7 , 8 ] , radar [ 9 ] , radio-frequency (RF) [ 10 , 11 , 12 , 13 , 14 ] signals and camera-based images and videos [ 15 , 16 ] that are used to detect and classify UAVs. Acoustic-based methodologies often exhibit heightened sensitivity to environmental noise, while the efficacy of visual data from cameras can be compromised by conditions such as obstructive structures, ambient lighting, and other local variables. In contrast, techniques that utilize RF signals and network traffic steam are generally less affected by such environmental changes. However, commercial UAVs typically operate on exclusive communication channels that are protected by various levels of encryption, presenting significant challenges in data acquisition and surveillance. Therefore, it is crucial to optimize the utilization of RF signals for UAV detection in environments susceptible to noise. Recent works have focused on applying zero-trust security mechanisms to enhance the security of UAV networks. For example, the authors in [ 17 ] proposed a zero-trust security framework for UAV networks, in detecting and identifying UAVs based on RF signals. Experimental results have demonstrated the effectiveness of the proposed zero-trust security framework in enhancing the security of UAV networks. Similarly, Ouiazzane et al. [ 18 ] developed a zero-trust security framework for UAV intrusion detection by using network traffic data. However, the existing zero-trust security mechanisms are not well integrated with federated learning, which may lead to potential security threats. Though there have been extensive works on UAV authentication and detection, the security and privacy of UAV networks are still challenging in three aspects, namely, task-level , data-level , and network-level security. • Firstly, existing task-level security methods rely on softmax classifiers trained for known UAVs, failing to identify unknown UAVs (UAV rejection). A robust mechanism for authenticating and rejecting UAVs based on unique features is essential. • Secondly, data-level security requires shifting from centralized deployment to distributed mechanisms that can authenticate and reject UAVs without sharing raw data, reducing single-point failure risks and enhancing network security. • Thirdly, network-level security remains neglected in existing UAV detection methods. A robust security mechanism is needed to protect the confidentiality, integrity, and availability of UAV-transmitted data against potential threats. Fig. 1: The methods of UAV detection and authentication. To advance UAV authentication in wireless networks, we propose a federated learning (FL)-based lightweight spectrogram network (LSNet) with zero trust for UAV authentication and rejection. The proposed model integrates the principles of zero trust security to protect wireless networks from potential security threats. The FL-based LSNet enables multiple clients to collaboratively train a global deep-learning model without sharing their raw data, thereby preserving the privacy of the data. Moreover, the FL-based LSNet with zero trust ensures that all model submission is authenticated, authorized, and encrypted, thereby minimizing the risk of security breaches. Then, to execute UAV authentication and rejection in a distributed manner, a lightweight spectrogram network with a novel loss function is proposed. The proposed LSNet can effectively authenticate and reject UAVs, thereby enhancing the security of UAV networks. The main contributions of this paper are summarized as follows 1. We propose a federated learning-based lightweight network with zero trust for enhancing the security of wireless networks. The framework integrates security at multiple levels simultaneously—task-level (distinguishing known vs. unknown UAVs), data-level (preserving privacy during model training), and network-level (protecting against external threats). This holistic approach provides significantly stronger security guarantees than systems focusing on only one security aspect. The framework enables multiple clients to collaboratively train a global deep-learning model without sharing their raw data, thereby preserving the privacy of the data. 2. To execute UAV authentication and rejection in a distributed manner, we present a lightweight spectrogram network (LSNet) with a novel loss function. Our experiments reveal that lightweight neural networks ( < 400K parameters) can achieve comparable or superior performance to much larger models for UAV authentication, with LSNet requiring only 1.6MB of storage. Multi-channel attention convolutional

Read the complete text