Article
Computer Science, Hardware & Architecture
Gianmarco Baldini, Raimondo Giuliani, Monica Gemo, Franc Dimc
Summary: The paper presents an alternative or complementary approach to identification and authentication based on physical properties of Hall sensors commonly used in vehicles. Different signal processing algorithms were applied to evaluate the experimental dataset, with dimensionality reduction using decimation filters to improve time efficiency while maintaining high accuracy in identification and authentication processes.
COMPUTERS & ELECTRICAL ENGINEERING
(2021)
Article
Biotechnology & Applied Microbiology
Edoardo Spairani, Beniamino Daniele, Maria Gabriella Signorini, Giovanni Magenes
Summary: This study proposes a hybrid approach based on a neural network for classifying healthy and pathological fetuses by combining quantitative parameters and image representations of the fetal heart rate signal. After training, the proposed neural network achieves an overall accuracy of 80.1% on a large dataset.
FRONTIERS IN BIOENGINEERING AND BIOTECHNOLOGY
(2022)
Article
Engineering, Electrical & Electronic
Shunjie Zhang, Tianhao Wu, Wei Wang, Ronghui Zhan, Jun Zhang
Summary: RF fingerprinting is a challenging and important technique for wireless device identification. Recent research has applied deep learning-based classifiers to ADS-B signals without losing aircraft ID information. However, traditional methods are not effective in achieving high accuracy for deep learning models to recognize RF signals. In this study, a Gaussian low-pass channel attention convolution network is proposed, which uses a Gaussian low-pass channel attention module (GLCAM) to extract low-frequency fingerprint features. Experimental results demonstrate that the method achieves an accuracy of 92.08%, which is 6.21% higher than convolutional neural networks.
ELECTRONICS LETTERS
(2023)
Article
Biology
Rui Hu, Jie Chen, Li Zhou
Summary: This paper proposes a novel transformer-based deep learning neural network, ECG DETR, for arrhythmia detection on continuous single-lead ECG segments. The model simultaneously predicts the positions and categories of all heartbeats within an ECG segment, eliminating the need for explicit heartbeat segmentation. The proposed method shows comparable performance to previous works, achieving high overall accuracy on different arrhythmia detection tasks.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Computer Science, Hardware & Architecture
Anu Jagannath, Jithin Jagannath, Prem Sagar Pattanshetty Vasanth Kumar
Summary: This paper surveys RF fingerprinting approaches, from traditional methods to the latest deep learning algorithms, comprehensively introducing the importance of RF fingerprinting technology in ensuring data privacy and security in wireless networks.
Article
Computer Science, Information Systems
Kemal Davaslioglu, Serdar Boztas, Mehmet Can Ertem, Yalin E. Sagduyu, Ender Ayanoglu
Summary: Deep learning has wide applications in the wireless domain for improving spectrum awareness. We propose a self-supervised RF signal representation learning method and apply it to automatic modulation recognition task. By capturing the wireless signal characteristics through a set of transformations, we demonstrate that self-supervised learning significantly increases the sample efficiency of automatic modulation recognition and maintains high accuracy even with limited training data.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2023)
Article
Chemistry, Analytical
Cameron J. Huggins, Rebecca Clarke, Daniel Abasolo, Erreka Gil-Rey, Jonathan H. Tobias, Kevin Deere, Sarah J. Allison
Summary: This study aims to evaluate the accuracy of machine learning models in binary and ternary classification tasks for postmenopausal women, and to more accurately characterize weight-bearing activities important for skeletal health using accelerometer data.
Article
Multidisciplinary Sciences
Christian Bergler, Simeon Q. Smeele, Stephen A. Tyndel, Alexander Barnhill, Sara T. Ortiz, Ammie K. Kalan, Rachael Xi Cheng, Signe Brinklov, Anna N. Osiecka, Jakob Tougaard, Freja Jakobsen, Magnus Wahlberg, Elmar Noeth, Andreas Maier, Barbara C. Klump
Summary: Bioacoustic research often relies on manual identification of target species or call types, which is time-consuming and error-prone. This study presents an open-source deep learning framework, ANIMAL-SPOT, which achieves high accuracy in identifying bioacoustic signals without requiring animal-specific machine learning approaches. The framework is accessible to a broad audience and does not rely on expert knowledge or special computing resources.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Information Systems
Aayesha, Muhammad Bilal Qureshi, Muhammad Afzaal, Muhammad Shuaib Qureshi, Muhammad Fayaz
Summary: This paper focuses on extracting distinguishing features of seizure EEG recordings to develop an approach that employs both fuzzy-based and traditional machine learning algorithms for epileptic seizure detection. The obtained results show that K-Nearest Neighbor (KNN) and Fuzzy Rough Nearest Neighbor (FRNN) give the highest classification accuracy scores, with improved sensitivity and specificity percentages.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Computer Science, Information Systems
Hao Chen, Seung-Jun Kim
Summary: Machine learning-based signal recognition algorithms are proposed in this work, utilizing discriminative dictionary learning algorithms with various feature-shaping constraints. The classifiers can robustly classify component signals even when a mixture of heterogeneous signal classes is observed, similar to multi-user detection in wireless communication. The algorithms are tested using real wideband RF measurement data, demonstrating their ability to classify signals effectively.
Article
Chemistry, Analytical
Paolo Brambilla, Chiara Conese, Davide Maria Fabris, Paolo Chiariotti, Marco Tarabini
Summary: Quality inspection in industrial production is benefiting from the combination of vision-based techniques and artificial intelligence algorithms. This paper discusses the defect identification problem for circularly symmetric mechanical components and compares the performances of a standard algorithm with a Deep Learning (DL) approach. The standard algorithm provides better results in terms of accuracy and computational time, but DL achieves high accuracy in identifying damaged teeth. The possibility of extending the methods and results to other circularly symmetrical components is also analyzed and discussed.
Review
Chemistry, Analytical
Lamia Alhoraibi, Daniyal Alghazzawi, Reemah Alhebshi, Osama Bassam J. Rabie
Summary: The physical layer security of wireless networks is crucial due to the rapid development of wireless communications and increasing security threats. However, there is scarce literature on physical layer authentication (PLA) and its role in improving wireless security. This paper aims to systematically compare existing studies on PLA and its applications, showcasing the latest techniques and suggesting future research directions. The study also explores the use of machine learning approaches for PLA in wireless communication systems, providing valuable insights for researchers and security model developers.
Article
Engineering, Biomedical
Ali Fatih Gunduz, Muhammed Fatih Talu
Summary: This study compares the detection and classification of persistent atrial fibrillation (PeAF) and paroxysmal atrial fibrillation (PAF) using deep learning models with different approaches. The results show that the classification approach based on spectral features achieves the highest training accuracy (0.9788), while the classification based on P wave detection achieves the highest test accuracy (0.8765). The bidirectional long short-term memory (BiLSTM) network outperforms the convolutional neural network (CNN) and LSTM cascades in capturing time-sensitive features of ECG signals.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Computer Science, Artificial Intelligence
Pasquale Foggia, Antonio Greco, Antonio Roberto, Alessia Saggese, Mario Vento
Summary: Current state-of-the-art audio analysis algorithms based on deep learning often use hand-crafted Spectrogram-like audio representations which have limitations. To address these limitations, we propose a new convolutional architecture called DEGramNet, trained with a learnable time-frequency representation called DEGram. The DEGramNet achieves state-of-the-art performance on the VGGSound dataset for sound event classification and comparable accuracy with a complex approach on the VoxCeleb dataset for speaker identification.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Zhongsen Sun, Kaizhuang Li, Yu Zheng, Xi Li, Yunlong Mao
Summary: Given the diversity of radar signals, this paper proposes an improved EfficientNetv2-s classification method based on deep learning for more precise classification and recognition of radar radiation source signals. The method uses 16 different types of radar signal parameters to generate random data sets consisting of spectrum images with varying amplitude. It replaces two-dimensional convolution in EfficientNetV2 with one-dimensional convolution and optimizes the channel attention mechanism to achieve better accuracy. Compared to other methods, this method has higher classification accuracy on the test set and lower complexity.
Article
Computer Science, Information Systems
Yalin E. Sagduyu, Yi Shi, Tugba Erpek
Summary: An adversarial deep learning approach is used to launch over-the-air spectrum poisoning attacks, where an adversary learns the behavior of a transmitter and falsifies the spectrum sensing data. The attacks are energy efficient, hard to detect, and substantially reduce throughput. A dynamic defense is designed to manipulate the adversary's training data and sustain the transmitter's throughput.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2021)
Article
Telecommunications
Yi Shi, Kemal Davaslioglu, Yalin E. Sagduyu
Summary: The paper presents a deep learning-based spoofing attack to bypass physical-layer signal authentication by generating synthetic wireless signals that cannot be statistically distinguished from intended transmissions. The attack increases the probability of misclassifying spoofing signals as intended signals by jointly capturing waveform, channel, and radio hardware effects inherent to wireless signals under attack. The success of the spoofing attack can be increased by the adversary transmitter using multiple antennas, while it decreases when the defender receiver uses multiple antennas.
IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING
(2021)
Article
Computer Science, Information Systems
Yi Shi, Yalin E. Sagduyu
Summary: This paper presents an over-the-air membership inference attack (MIA) that can leak private information from a wireless signal classifier. The attack uses machine learning to classify wireless signals, which is useful for PHY-layer authentication. The MIA infers whether a signal has been used in the training data of a target classifier and can exploit the leaked information to identify vulnerabilities. The paper also proposes a proactive defense strategy against the MIA, which involves building a shadow model to deceive the adversary and reduce the accuracy of the attack.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Computer Science, Information Systems
Zhengping Luo, Shangqing Zhao, Zhuo Lu, Jie Xu, Yalin E. Sagduyu
Summary: This paper proposes a novel adversarial machine learning framework called Learning-Evaluation-Beating (LEB) to mislead the fusion center in cooperative spectrum sensing. The LEB attack effectively beats existing defense strategies by creating malicious sensing data using a surrogate model. Additionally, a non-invasive defense method named influence-limiting defense is introduced to reduce the overall disruption caused by LEB attack.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2022)
Article
Physics, Multidisciplinary
Brian Kim, Yalin Sagduyu, Kemal Davaslioglu, Tugba Erpek, Sennur Ulukus
Summary: This paper investigates the privacy of wireless communications against eavesdroppers who use a deep learning classifier to detect transmissions of interest. By utilizing a cooperative jammer that sends adversarial perturbations, the eavesdropper can be deceived into misclassifying the received signals as noise. The results show that the adversarial perturbation is more effective when multiple antennas are used.
Article
Engineering, Electrical & Electronic
Brian Kim, Yalin E. Sagduyu, Kemal Davaslioglu, Tugba Erpek, Sennur Ulukus
Summary: This paper introduces channel-aware adversarial attacks against deep learning-based wireless signal classifiers, exploring the failure of evasion attacks without considering channel effects and presenting realistic attacks by considering channel effects. It also proposes a certified defense based on randomized smoothing to make the modulation classifier robust to adversarial attacks.
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS
(2022)
Article
Telecommunications
Ender Ayanoglu, Kemal Davaslioglu, Yalin E. Sagduyu
Summary: This paper investigates the application of Generative Adversarial Networks (GANs) in next-generation communications, particularly in addressing spectrum sharing, anomaly detection, and security attack mitigation in cognitive networks. GANs have advantages such as learning and synthesizing field data, pre-training classifiers, increasing resolution, and recovering corrupted bits.
IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING
(2022)
Article
Computer Science, Information Systems
Xueyuan Wang, M. Cenk Gursoy, Tugba Erpek, Yalin E. Sagduyu
Summary: This article discusses the challenging task of determining collision-free trajectory in multi-UAV noncooperative scenarios while collecting data from distributed IoT nodes. The problem is translated into a Markov decision process and a Dueling double deep Q-network algorithm is proposed to learn the decision-making policy. Numerical results demonstrate efficient real-time navigation in various scenarios.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Computer Science, Information Systems
Kemal Davaslioglu, Serdar Boztas, Mehmet Can Ertem, Yalin E. Sagduyu, Ender Ayanoglu
Summary: Deep learning has wide applications in the wireless domain for improving spectrum awareness. We propose a self-supervised RF signal representation learning method and apply it to automatic modulation recognition task. By capturing the wireless signal characteristics through a set of transformations, we demonstrate that self-supervised learning significantly increases the sample efficiency of automatic modulation recognition and maintains high accuracy even with limited training data.
IEEE WIRELESS COMMUNICATIONS LETTERS
(2023)
Article
Computer Science, Information Systems
Yaliang Shi, Qiuling Yang, Xiwen Huang, Deshun Li, Xiangdang Huang
Summary: This article proposes a load-balanced and QoS-aware software-defined IoUT framework using the SDN+AI paradigm. By adopting SDN technology to separate the data plane and control plane, the network's scalability and flexibility are enhanced. The CASM load-balancing strategy and SQAR adaptive routing protocol based on reinforcement learning further optimize the network's performance in terms of load balancing and QoS satisfaction. Experimental results show that CASM achieves efficient load balancing and SQAR outperforms existing QoS-aware routing protocols. Overall, the proposed framework maintains a low QoS violation rate and high load-balancing rate in a timely manner.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Review
Computer Science, Information Systems
Damilola Adesina, Chung-Chu Hsieh, Yalin E. Sagduyu, Lijun Qian
Summary: Machine learning (ML) is effective in learning from spectrum data and solving complex wireless communication tasks, and deep learning (DL) has found success in various wireless communication tasks. However, ML and DL are vulnerable to attacks, leading to the field of adversarial machine learning (AML). AML in the wireless communications domain is still in its early stage. This paper presents a comprehensive review of the latest research efforts focused on AML in wireless communications, discussing AML attacks, defense mechanisms, and future outlook.
IEEE COMMUNICATIONS SURVEYS AND TUTORIALS
(2023)
Article
Engineering, Electrical & Electronic
Yi Shi, Yalin E. Sagduyu, Tugba Erpek, M. Cenk Gursoy
Summary: In this paper, the use of reinforcement learning (RL) for network slicing in NextG radio access networks is explored. An over-the-air attack is introduced based on adversarial machine learning to manipulate the RL algorithm and disrupt network slicing. The attack is shown to be more effective than benchmark jamming attacks. Different defense schemes are introduced to defend against this attack and improve the RL algorithm's reward.
IEEE OPEN JOURNAL OF VEHICULAR TECHNOLOGY
(2023)
