PCA-Featured Transformer for Jamming Detection in 5G UAV Networks
Joseanne Viana, Hamed Farkhari, Pedro Sebastião, Victor P Gil Jimenez · 2024
Read PCA-Featured Transformer for Jamming Detection in 5G UAV Networks by Joseanne Viana, Hamed Farkhari, Pedro Sebastião, Victor P Gil Jimenez (2024) 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-2412-15312.txt.
Introduction
Joseanne Viana, Hamed Farkhari, Pedro Sebastião, and Victor P. Gil Jimenez aim to spot jamming attacks on drones linked by fifth-generation networks. Their paper from two thousand twenty-four warns that smart jammers change their tricks fast. Old detectors miss them, and many learning methods need heavy feature work or read time patterns poorly.
Their design is a U-shaped transformer that first shrinks signal features with a classic stats tool. Training uses an entropy term in the loss, an idea borrowed from a control method. Custom position coding tracks the repeating shape of wireless signals. In tests, the system caught most attacks even with no direct line of sight, and trained far faster.
Catalog record
| CREATOR | Joseanne Viana, Hamed Farkhari, Pedro Sebastião, Victor P Gil Jimenez |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-15312
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
PCA-Featured Transformer for Jamming Detection in 5G UAV Networks Joseanne Viana, Hamed Farkhari, Pedro Sebastião, Victor P Gil Jimenez Abstract Unmanned Aerial Vehicles (UAVs) face significant security risks from jamming attacks, which can compromise network functionality. Traditional detection methods often fall short when confronting AI-powered jamming that dynamically modifies its behavior, while contemporary machine learning approaches frequently demand substantial feature engineering and struggle with temporal patterns in attack signatures. The vulnerability extends to 5G networks employing Time Division Duplex (TDD) or Frequency Division Duplex (FDD), where service quality may deteriorate due to deliberate interference. We introduce a novel U-shaped transformer architecture that leverages Principal Component Analysis (PCA) to refine feature representations for improved wireless security. The training process is regularized by incorporating the output entropy uncertainty into the loss function, a mechanism inspired by the Soft Actor-Critic (SAC) algorithm in Reinforcement Learning (RL) to enable robust jamming detection techniques. The architecture features a modified transformer encoder specially designed to process critical wireless signal features, including Received Signal Strength Indicator (RSSI) and Signal-to-Interference-plus-Noise Ratio (SINR) measurements. We complement this with a custom positional encoding mechanism that specifically accounts for the inherent periodicity of wireless signals, enabling a more accurate representation of temporal signal patterns. In addition, we propose a batch size scheduler and implement chunking techniques to optimize training convergence for time series data. These advancements contribute to achieving up to a ten times improvement in training speed within the advanced U-shaped encoder-decoder transformer model introduced in this study. Experimental evaluations demonstrate the effectiveness of our entropy-based approach, achieving detection rates of 89.46% under Line-of-Sight (LoS) conditions and 85.06% in non-Line-of-Sight (NLoS) scenarios. Our method significantly outperforms existing solutions, surpassing XGBoost (XGB) classifiers by approximately 4.5% and other deep learning approach by more than 2%. † † history: Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000. † † doi: 10.1109/ACCESS.2023.0322000 † † address: UC3M - Universidad Carlos III de Madrid, Madrid, Spain; † † address: ISCTE – Instituto Universitário de Lisboa, Av. das Forças Armadas, 1649-026 Lisbon, Portugal; † † address: Tyndall National Institute, Ireland; † † address: IT – Instituto de Telecomunicações, Av. Rovisco Pais, 1, Torre Norte, Piso 10, 1049-001 Lisboa, Portugal; † † corresponding: Corresponding author: Joseanne Viana ([email protected]) § § footnotetext: Collaborative authors with equal contribution I Introduction The intersection of Unmanned Aerial Vehicles (UAVs) and wireless communication systems represents a rapidly evolving research domain with significant technological and security implications. As UAVs transition from specialized military applications to widespread commercial deployment, their integration into communication networks introduces novel challenges and opportunities [ 1 ] , [ 2 ] , [ 3 ] , [ 4 ] , [ 5 ] , [ 6 ] . Base stations mounted on UAVs demonstrate potential for applications including emergency response, surveillance of borders, and providing temporary network coverage [ 7 ] , [ 8 ] , [ 9 ] , [ 4 ] . However, when UAVs function as end-devices in services such as package delivery, they introduce distinct security vulnerabilities within 5G networks [ 10 ] , [ 11 ] . A critical concern is the vulnerability of UAV communication systems to sophisticated jamming attacks, which can manipulate Time Division Duplex (TDD) and Frequency Division Duplex (FDD) systems, resulting in severe service disruptions with impacts reaching up to 99% in TDD uplink and 82% in FDD downlink scenarios [ 12 ] , [ 13 ] , [ 14 ] . Jamming can affect private networks, compromising the integrity and availability of mission critical communications in industrial, corporate, and specialized operational environments where UAVs are increasingly deployed. Machine learning, particularly deep learning approaches, offers promising avenues for developing more effective and proactive jamming detection systems [ 15 ] , [ 16 ] , [ 17 ] . Convolutional Neural Networks (CNNs) have shown effectiveness in extracting spatial features from signal data [ 18 ] , [ 19 ] , [ 20 ] , while Long Short-Term Memory (LSTM) networks excel at modeling temporal patterns in wireless communications [ 21 ] , [ 22 ] , [ 23 ] , [ 24 ] . Several studies have advanced this field. [ 25 ] developed a branched deep neural network architecture for simultaneous jamming detection and link scheduling in dense wireless networks. Their system employs two specialized subnetworks: one leverages geographical information and signal power measurements to detect and locate jammers, while the other optimizes link scheduling based on jamming detection results to maximize network throughput under adverse conditions. [ 26 ] proposed neural networks for compound jamming signal recognition, while [ 27 ] introduced a singular value decomposition approach for jamming identification in Global Navigation Satellite System (GNSS)-based systems. For radar systems specifically, [ 28 ] introduced a deep learning method to counter interrupted sampling deceptive jamming. Their four-stage approach begins by generating time-frequency representations of radar echoes using short-time Fourier transform, which feed into a You Only Look Once (YOLO) object detection model that identifies jammed signals while implementing position correction to preserve legitimate target information. [ 29 ] further expanded this domain by proposing deep neural networks for digital radio frequency memory (DRFM) jamming mode identification. Addressing cognitive radio vulnerabilities, [ 30 ] created a one-dimensional convolutional neural network operating directly on raw signal data to detect primary user emulation and jamming attacks. This approach eliminates the manual feature engineering required by traditional methods, as noted by [ 31 ] . Their architecture incorporates three convolutional layers with rectified linear unit (ReLU) activation functions, followed by dense and softmax output layers. More recently, [ 32 ] developed a deep learning system using ensemble techniques for detecting jamming attacks in 5G networks, combining RF domain and physical layer features with a Temporal Epistemic Decision Aggregator to enhance detection reliability despite signal impairments and carrier frequency offset. Similarly, [ 33 ] proposed a feature- and spectrogram-tailored machine learning approach for jamming detection in OFDM-based UAVs. Despite these advances, these architectures often struggle to capture the complex interdependencies and multifaceted patterns characteristic of modern jamming attacks, particularly in heterogeneous environments that incorporate both 5G New Radio (NR) and Narrowband Internet of Things (NB-IoT) interfaces [ 34 ] , [ 35 ] , [ 13 ] , [ 36 ] . Transformer architectures, with their self-attention mechanisms [ 37 ] , have improved sequential data analysis and offer significant potential for jamming detection [ 38 ] . These models are particularly well-suited for identifying long-range dependencies in wireless signal data, enabling the detection of sophisticated jamming patterns across various time scales and frequency bands [ 10654316 ] . The multi-head attention mechanism further enhances this capability by simultaneously analyzing diverse signal attributes, from immediate interference to subtle, persistent disruptions, making transformers especially effective against energy-efficient selective jamming techniques [ 39 ] . With the accelerating deployment of 5G networks and increasingly complex security threats [ 40 ] , [ 41 ] , there is a need for advanced jamming detection methods that can preemptively identify and counteract these attacks [ 42 ] , [ 43 ] . Integrating transformers with other machine learning techniques offers promising new approaches for addressing these challenges, establishing foundations for more resilient