Marwan Krunz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Marwan Krunz has a total of 133 co-authors. Here are the top co-authors by number of shared publications:
- Yong Xiao
School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan, China
16 shared publications - Diep N. Nguyen
School of Electrical and Data Engineering, University of Technology Sydney, Ultimo, NSW, Australia
12 shared publications - Dinh Thai Hoang
University of Technology Sydney
7 shared publications - Guangming Shi
Peng Cheng Laboratory, Shenzhen, Guangdong, China
7 shared publications - Eryk Dutkiewicz
School of Electrical and Data Engineering, University of Technology Sydney, Sydney, NSW, Australia
6 shared publications - Mingjie Feng
Wuhan National Laboratory for Optoelectronics, Huazhong University of Science and Technology, Wuhan, China
4 shared publications
Latest Publications
- Open Access
- 22 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 101-110
Physics-Informed Kernelized Thompson Sampling for Beam Management in Millimeter-Wave Systems
- 22 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 101-110
Open AccessPhysics-Informed Kernelized Thompson Sampling for Beam Management in Millimeter-Wave Systems
- Open Access
- 30 October 2026
- Published by EDP Sciences in Astronomy & Astrophysics
- Vol. 714, A505
Weakly supervised machine learning for enhanced radio frequency interference detection in astronomical observations
Radio astronomical observations are prone to radio frequency interference (RFI) from manmade signals, including cellular transmissions, satellites, automotive radar, and others. Achieving harmonious spectrum sharing between the radio astronomy (RA) service and wireless systems requires identifying RFI in the vicinity of an RA telescope in real time. In this paper, we propose a weakly supervised generative adversarial network (GAN) for RFI detection in astronomical observations. Weakly supervised training is particularly effective when dealing with largely unlabeled datasets, as is the case with RA observations. At the same time, adversarial training through a GAN can enhance RFI detection under class imbalance. Our study is based on radio astronomical scans obtained from two telescopes: the 12-meter telescope at Kitt Peak, Arizona, operated by the Arizona Radio Observatory (ARO), and the Low-Frequency Array (LOFAR) telescope in the Netherlands. We further complemented these scans with a simulated dataset generated for the Hydrogen Epoch of Reionization Array (HERA) telescope. For the ARO dataset, the scans represent the measured temperature (equivalently, power density) of the telescope’s antenna versus sky frequency over a 4 GHz bandwidth. We split each scan into 200 frequency channels (20 MHz per channel). A very small fraction of the dataset is manually labeled as clean or dirty (if RFI is present). We trained a reference binary classifier using only the manually labeled dataset. We then designed and trained two weakly supervised generative models: a semi-supervised GAN and an information maximization GAN. These models leverage the latent representations learned from the unlabeled data to improve the classifier’s F1 score and true positive rate (TPR). Next, we considered the LOFAR and HERA datasets and applied our two GAN methods on the spectrograms obtained from these datasets. Each spectrogram in the LOFAR dataset covers a frequency range of 120–190 MHz and a 10-minute time duration, whereas each HERA spectrogram covers a 105–195 MHz frequency band over a 30-minute observation period. We partitioned each spectrogram into smaller patches, which serve as inputs for RFI detection under both generative learning models. Our weakly supervised approaches significantly enhance the classification performance on all the ARO, LOFAR, and HERA datasets, achieving a TPR of 0.95, 0.79, and 1 respectively, compared to 0.65, 0, and 0 when such approaches are not used.
- 30 October 2026
- Published by EDP Sciences in Astronomy & Astrophysics
- Vol. 714, A505
Open AccessWeakly supervised machine learning for enhanced radio frequency interference detection in astronomical observations
Radio astronomical observations are prone to radio frequency interference (RFI) from manmade signals, including cellular transmissions, satellites, automotive radar, and others. Achieving harmonious spectrum sharing between the radio astronomy (RA) service and wireless systems requires identifying RFI in the vicinity of an RA telescope in real time. In this paper, we propose a weakly supervised generative adversarial network (GAN) for RFI detection in astronomical observations. Weakly supervised training is particularly effective when dealing with largely unlabeled datasets, as is the case with RA observations. At the same time, adversarial training through a GAN can enhance RFI detection under class imbalance. Our study is based on radio astronomical scans obtained from two telescopes: the 12-meter telescope at Kitt Peak, Arizona, operated by the Arizona Radio Observatory (ARO), and the Low-Frequency Array (LOFAR) telescope in the Netherlands. We further complemented these scans with a simulated dataset generated for the Hydrogen Epoch of Reionization Array (HERA) telescope. For the ARO dataset, the scans represent the measured temperature (equivalently, power density) of the telescope’s antenna versus sky frequency over a 4 GHz bandwidth. We split each scan into 200 frequency channels (20 MHz per channel). A very small fraction of the dataset is manually labeled as clean or dirty (if RFI is present). We trained a reference binary classifier using only the manually labeled dataset. We then designed and trained two weakly supervised generative models: a semi-supervised GAN and an information maximization GAN. These models leverage the latent representations learned from the unlabeled data to improve the classifier’s F1 score and true positive rate (TPR). Next, we considered the LOFAR and HERA datasets and applied our two GAN methods on the spectrograms obtained from these datasets. Each spectrogram in the LOFAR dataset covers a frequency range of 120–190 MHz and a 10-minute time duration, whereas each HERA spectrogram covers a 105–195 MHz frequency band over a 30-minute observation period. We partitioned each spectrogram into smaller patches, which serve as inputs for RFI detection under both generative learning models. Our weakly supervised approaches significantly enhance the classification performance on all the ARO, LOFAR, and HERA datasets, achieving a TPR of 0.95, 0.79, and 1 respectively, compared to 0.65, 0, and 0 when such approaches are not used.
- Open Access
- 15 August 2026
- Published by Springer Nature in Scientific Reports
SynCellNet: generative AI framework for single-cell RNA sequencing data generation and validation
- 15 August 2026
- Published by Springer Nature in Scientific Reports
Open AccessSynCellNet: generative AI framework for single-cell RNA sequencing data generation and validation
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (7) , 9962-9979
Preamble Signaling in WLAN: Vulnerabilities, Attacks, and Countermeasures
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (7) , 9962-9979
Preamble Signaling in WLAN: Vulnerabilities, Attacks, and Countermeasures
- Open Access
- 23 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Machine Learning in Communications and Networking
- Vol. 4, 612-628
ARCS: Adversarial RF Map Categorization and Synthesis Through Unsupervised Feature Extraction
In wireless networks, radio-frequency coverage maps (RF maps) are critical for tasks such as capacity planning, coverage estimation, and localization. Traditional methods for obtaining these maps, including site surveys and ray-tracing simulations, are either labor-intensive or computationally expensive, particularly at high frequencies. Generative AI offers a promising alternative for RF map synthesis and data augmentation. However, supervised generative approaches are often infeasible due to the lack of labeled training data, while unsupervised methods typically lack control over the generation process, limiting their practical utility. To overcome these challenges, we propose ARCS (Adversarial RF Map Categorization and Synthesis), a novel generative adversarial network (GAN)-based framework for unsupervised RF map categorization and synthesis. ARCS leverages the principles of information maximizing GAN (InfoGAN) to learn the latent structure of RF maps in an unsupervised manner during training and, through manipulation of the learned and interpretable latent codes, enables controlled generation of RF maps across floor-plan regions and Tx locations during inference. To improve training stability and synthesis quality, we integrate a gradient penalty-based Wasserstein GAN objective function along with a customized gradient-based loss function. Extensive experiments on both experimental and simulated datasets show that ARCS generates high-quality RF maps, associates them with discrete regions of the floor plan, and provides fine-grained control of Tx location within each region. Compared with a UNet-based conditional GAN and a conditional diffusion model, ARCS attains the best scores across structural similarity index metric (SSIM), PSNR, MAPE, RMSE, cosine similarity (CS), and Jensen–Shannon divergence (JSD). Moreover, ARCS is extremely fast, requiring a few milliseconds per synthetic map, compared to over 14 seconds with ray-tracing.
- 23 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Machine Learning in Communications and Networking
- Vol. 4, 612-628
Open AccessARCS: Adversarial RF Map Categorization and Synthesis Through Unsupervised Feature Extraction
In wireless networks, radio-frequency coverage maps (RF maps) are critical for tasks such as capacity planning, coverage estimation, and localization. Traditional methods for obtaining these maps, including site surveys and ray-tracing simulations, are either labor-intensive or computationally expensive, particularly at high frequencies. Generative AI offers a promising alternative for RF map synthesis and data augmentation. However, supervised generative approaches are often infeasible due to the lack of labeled training data, while unsupervised methods typically lack control over the generation process, limiting their practical utility. To overcome these challenges, we propose ARCS (Adversarial RF Map Categorization and Synthesis), a novel generative adversarial network (GAN)-based framework for unsupervised RF map categorization and synthesis. ARCS leverages the principles of information maximizing GAN (InfoGAN) to learn the latent structure of RF maps in an unsupervised manner during training and, through manipulation of the learned and interpretable latent codes, enables controlled generation of RF maps across floor-plan regions and Tx locations during inference. To improve training stability and synthesis quality, we integrate a gradient penalty-based Wasserstein GAN objective function along with a customized gradient-based loss function. Extensive experiments on both experimental and simulated datasets show that ARCS generates high-quality RF maps, associates them with discrete regions of the floor plan, and provides fine-grained control of Tx location within each region. Compared with a UNet-based conditional GAN and a conditional diffusion model, ARCS attains the best scores across structural similarity index metric (SSIM), PSNR, MAPE, RMSE, cosine similarity (CS), and Jensen–Shannon divergence (JSD). Moreover, ARCS is extremely fast, requiring a few milliseconds per synthetic map, compared to over 14 seconds with ray-tracing.
- 18 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 74 (00906778) , 6332-6348
Deep Learning-Driven Friendly Jamming for Secure Multicarrier ISAC Under Channel Uncertainty
Integrated sensing and communication (ISAC) systems promise efficient spectrum utilization by jointly supporting radar sensing and wireless communication. This paper presents a deep learning-driven framework for enhancing physical-layer security in multicarrier ISAC systems under imperfect channel state information (CSI) and in the presence of unknown eaves-dropper (Eve) locations. Unlike conventional ISAC-based friendly jamming (FJ) approaches that require Eve’s CSI or precise angle-of-arrival (AoA) estimates, our method exploits radar echo feedback to guide directional jamming without explicit Eve’s information. To enhance robustness to radar sensing uncertainty, we propose a radar-aware neural network that jointly optimizes beamforming and jamming by integrating a novel nonparametric Fisher Information Matrix (FIM) estimator based on f-divergence. The jamming design satisfies the Cramér–Rao lower bound (CRLB) constraints even in the presence of noisy AoA. For efficient implementation, we introduce a quantized tensor train-based encoder that reduces the model size by more than 100 times with negligible performance loss. We also integrate a non-overlapping secure scheme into the proposed framework, in which specific sub-bands can be dedicated solely to communication. Extensive simulations demonstrate that the proposed solution achieves significant improvements in secrecy rate, reduced block error rate (BLER), and strong robustness against CSI uncertainty and angular estimation errors, under-scoring the effectiveness of the proposed deep learning–driven friendly jamming framework under practical ISAC impairments.
- 18 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Communications
- Vol. 74 (00906778) , 6332-6348
Deep Learning-Driven Friendly Jamming for Secure Multicarrier ISAC Under Channel Uncertainty
Integrated sensing and communication (ISAC) systems promise efficient spectrum utilization by jointly supporting radar sensing and wireless communication. This paper presents a deep learning-driven framework for enhancing physical-layer security in multicarrier ISAC systems under imperfect channel state information (CSI) and in the presence of unknown eaves-dropper (Eve) locations. Unlike conventional ISAC-based friendly jamming (FJ) approaches that require Eve’s CSI or precise angle-of-arrival (AoA) estimates, our method exploits radar echo feedback to guide directional jamming without explicit Eve’s information. To enhance robustness to radar sensing uncertainty, we propose a radar-aware neural network that jointly optimizes beamforming and jamming by integrating a novel nonparametric Fisher Information Matrix (FIM) estimator based on f-divergence. The jamming design satisfies the Cramér–Rao lower bound (CRLB) constraints even in the presence of noisy AoA. For efficient implementation, we introduce a quantized tensor train-based encoder that reduces the model size by more than 100 times with negligible performance loss. We also integrate a non-overlapping secure scheme into the proposed framework, in which specific sub-bands can be dedicated solely to communication. Extensive simulations demonstrate that the proposed solution achieves significant improvements in secrecy rate, reduced block error rate (BLER), and strong robustness against CSI uncertainty and angular estimation errors, under-scoring the effectiveness of the proposed deep learning–driven friendly jamming framework under practical ISAC impairments.
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Networking
- Vol. 34, 3123-3136
Online Learning for Edge Node Program Placement in Mobile Edge Computing Networks
Mobile edge computing (MEC) is a key technology to support computationally intensive mobile applications with stringent latency requirements. With MEC servers deployed at network edge (e.g., base stations), the computational tasks generated by various applications can be offloaded to nearby edge nodes (ENs) and timely processed there. Meanwhile, future mobile applications will be more diverse and complex, which will need to be supported by a large number of complicated programs. As the storage space of ENs is limited, it is infeasible for each EN to store the program codes of all applications. Thus, it is necessary to optimize program placement at ENs to fully harvest the potential of MEC. In this paper, we investigate the problem of program placement and user association in storage-limited MEC networks. Such a problem is formulated as a sequential decision-making problem. We first consider the single EN scenario and propose an online learning-based solution. We then propose a solution framework for the multi-EN scenario, where we decompose the original problem into three subproblems and iteratively solve them with low-complexity approaches. Simulation results show that the average latency achieved by our proposed schemes is 30% to 70% lower than two benchmark schemes and is on average less than 10% higher than a lower bound.
- 28 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Networking
- Vol. 34, 3123-3136
Online Learning for Edge Node Program Placement in Mobile Edge Computing Networks
Mobile edge computing (MEC) is a key technology to support computationally intensive mobile applications with stringent latency requirements. With MEC servers deployed at network edge (e.g., base stations), the computational tasks generated by various applications can be offloaded to nearby edge nodes (ENs) and timely processed there. Meanwhile, future mobile applications will be more diverse and complex, which will need to be supported by a large number of complicated programs. As the storage space of ENs is limited, it is infeasible for each EN to store the program codes of all applications. Thus, it is necessary to optimize program placement at ENs to fully harvest the potential of MEC. In this paper, we investigate the problem of program placement and user association in storage-limited MEC networks. Such a problem is formulated as a sequential decision-making problem. We first consider the single EN scenario and propose an online learning-based solution. We then propose a solution framework for the multi-EN scenario, where we decompose the original problem into three subproblems and iteratively solve them with low-complexity approaches. Simulation results show that the average latency achieved by our proposed schemes is 30% to 70% lower than two benchmark schemes and is on average less than 10% higher than a lower bound.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (9) , 1-18
SANet: A Semantic-aware Agentic AI Networking Framework for Cross-layer Optimization in 6G
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (9) , 1-18
SANet: A Semantic-aware Agentic AI Networking Framework for Cross-layer Optimization in 6G