2026. 2^nd Issue

Volume XVIII, Number 2

Table of contents 

Full issue  

PAPERS FROM OPEN CALL

Zoltán Szatmáry, Balázs Szenczy, and Tamás Lévai
Efficient Media Routing for Media over QUIC: Balancing Cost and Latency 

Media accounts for over 80% of global Internet traffic, much of it being latency-sensitive. The IETF’s Media over QUIC (MoQ) protocol draft leverages QUIC for calable media delivery, using a relay-based architecture similar to CDNs. However, current MoQ implementations rely on full mesh relay networks, leading to high network utilization and scalability issues. In this work, we present a novel approach to enforce media routing and lower the overall costs of operating a MoQ relay topology while satisfying delay service level objectives. We formulate the problem mathematically, analyze its complexity, and develop optimal and heuristic algorithms. We integrate our approach with a widely-used MoQ implementation. Our evaluation using different synthetic and realistic topologies shows improvements in cost efficiency while keeping all delay constraints satisfied.

DOI: 10.36244/ICJ.2026.2.1
Download 

José Gómez-delaHiz, Juan Luis Herrera, Sergio Laso, Javier BerrocalJaime Galán-Jiménez
Emulating the SDN-Enabled Computing Continuum through Lightweight Virtualization 

In recent years, the number of Internet-connected devices has increased notably, leading to a significant rise in data traffic. This increase has been fostered by the Internet of Things paradigm, the use of microservices architectures in application development, and the ability to deploy these applications across various layers in the Computing Continuum (including Fog, Edge, and Cloud layers). Consequently, choosing the right deployment strategy has become essential for network operators and developers, especially in intensive domains such as smart cities. In this work, we introduce an emulation framework that allows developers and operators to decide how to deploy networks, computing devices, and applications in a Computing Continuum environment, ensuring compliance with established Quality of Service standards. This framework supports both IP and SDN network paradigms and is highly adaptable to different scenarios due to its use of container-based virtualization. Furthermore, the SDN paradigm provides flexibility, enabling the implementation of a service discovery feature that simplifies communication between end devices and services. Evaluations conducted in a realistic smart city scenario show that this framework can be extended and applied to a wide range of situations and configurations, meeting the needs of the research community in the Computing Continuum domain.

DOI: 10.36244/ICJ.2026.2.2
 Download 

Péter Baranyi, and Ádám B. Csapo
Introducing Neural Mesh for Logical Synthesis Models 

The paper introduces the Neural Mesh, a novel architecture that serves as an alternative to conventional neural network models widely used in contemporary AI systems. In parallel, the paper introduces the term Logical Synthesis Model (LSM) as a complement to Large Language Models (LLMs). While LLMs primarily rely on statistical representations and language-based reasoning to acquire and generate knowledge, Neural Mesh-based LSMs focus on structured representations and engineering reasoning to represent, analyze, synthesize, and control systems. LLMs excel at learning and reasoning over human knowledge expressed through language, whereas LSMs aim to learn structured representations that support the “understanding” of physical systems. In this sense, LLMs and LSMs may play roles analogous to the cerebrum and the cerebellum in biological intelligence. The novelty of the paper lies in bridging and combining the concepts of TP models, TP model transformation, and neural-network architectures. The proposed Neural Mesh is mathematically equivalent to the well-established TP model function family. The TP model transfor-mation provides a framework for constructing TP models with unique features that are particularly advantageous for systems and control applications. The contribution of the Neural Mesh is that it represents these unique features in the form of a special neural-network architecture, where these features are expressed as trainable neural connections and parameters. As a result, elements that are traditionally determined through an offline TP model transformation become directly tunable through learn-ing, while preserving the mathematical structure and control-theoretic advantages of the TP model framework. Therefore, the Neural Mesh representation opens a future research direction in which the TP model transformation is effectively embedded into the training process itself. Rather than generating a TP model through a subsequent offline transformation, the corresponding TP-model components are directly constructed and tuned during system identification via neural-network training tools.

DOI: 10.36244/ICJ.2026.2.3
 Download 

Salwa Marwan Salih, and Shelan Khasro Tawfeeq
Design and Numerical Analysis of 45-degree Polarization Rotator for Quantum Cryptography 

Polarization manipulation in visible wavelengths has an important role in many applications, such as free-space quantum communication and quantum cryptography. In this study, a polarization rotator based on a tilted slot waveguide by 45-degree is proposed and designed. Its principle is to rotate a +45°/-45° input polarized light to TM0/TE0 output polarized light. The simulation results show an extinction ratio of 45.23 (41.17) dB and insertion loss below 0.62 dB at 700 nm wavelength. The bandwidth for an extinction ratio higher than 20 dB is 40 nm. In addition, the device tolerance is analyzed.

DOI: 10.36244/ICJ.2026.2.4
Download 

Chuanchuan Teng, and Xizhuo Han
Supply Chain Logistics Demand Modeling and Abnormal Warning Based on Autoformer 

Current methods have limited capabilities in accurately modeling and dynamically warning of anomalies in port supply chain logistics (SCLs) scenarios where long-term trends in logistics demand (LD) coexist with short-term disturbances. To improve the accuracy of LD forecasting and anomaly warning (AW), and enhance the stability and responsiveness of the supply chain (SC), this paper constructed a multi-scale demand forecasting and AW model based on Autoformer. First, multi-source logistics data was integrated and combined with the Autoformer trend-disturbance decomposition mechanism to achieve structured modeling of non-stationary demand sequences. Then, a multi-scale attention path was designed to extract periodic information at the daily, monthly, and quarterly levels, respectively, to enhance the model's perception of multi-frequency features. Finally, a residual-driven anomaly scoring function and a dynamic threshold strategy were combined to construct a multi-level AW mechanism to achieve accurate identification of logistics risk events of different intensities. Experimental results show that in LD modeling, the Autoformer achieves a final weighted average percentage wrror (WAPE) and root mean square error (RMSE) of 0.105, and a matching degree of over 0.900 for the changing trend of supply chain logistics demand (SCLD) for a 200 twenty-feet equivalent unit (TEU) SC. In terms of AW, the Autoformer achieves average false negative rate (FNR) and false positive rate (FPR) of 12.0% and 7.3%, respectively, with an average overall delay of 1.3 hours. The weighted Kappa coefficient for consist-ency in AW levels is 0.844. The results show that the proposed Autoformer model, on the port logistics dataset used, demonstrates certain improvements in demand forecasting and anomaly detection compared to the selected baseline model, providing a feasible solution reference for forecasting and early warning tasks in intelligent supply chain management.

DOI: 10.36244/ICJ.2026.2.5
Download 

Mohammed Imad Aal-Nouman, Sarmad M. Hadi and Haider A. H. Alobaidy
Federated Reinforcement Learning for Load Balancing in O-RAN 5G and Beyond Networks 

The intelligent distributed control is a critical factor in next-generation mobile networks for connecting a massive number of devices and handling diverse services. The Open Radio Access Network (O-RAN) has recently gained significant traction as a flexible architecture based on standards that enables disaggregated, programmable network components. However, because O-RAN is inherently distributed, it creates an imbalance in traffic load when user mobility and traffic are dynamic. This paper proposes a Federated Reinforcement Learning (FedRL) framework to help load balance O- RAN units (O-RUs). The framework enables each O-RU to train a local reinforcement learning agent using real-time information on signal quality, mobility dynamics, and load conditions. The trained agents periodically share their learned information via federated averaging of their Q-tables at the Near-Real- Time RAN Intelligent Controller (Near- RT RIC), enabling collaborative learning without requiring centralized data collection. Using a synthetic dataset obtained from MATLAB-generated simulations of a mobile network of 500 m x 500 m, the proposed system is trained and tested. It simulates realistic user movement and radio conditions between multiple cells. In the experiments, FedRL reduces load imbalance by over 50% compared to rule-based schemes and by around 25% compared to centralized RL. It also achieves fairness values exceeding 0.95, increases throughput by more than 15%, increases handover success rates by around 13%, and reduces ping-pong events by 30-40 % compared to the baselines.

DOI: 10.36244/ICJ.2026.2.6
Download 

Szilárd László Takács, and Konrád Kajdy
Convolutional and Variational Autoencoders with Supervised Classifiers for RF Spectrum Anomaly Detection 

Effective and secure monitoring of the radio frequency spectrum is essential for the reliable operation of modern wireless communication systems. Automated detection of spectral anomalies can support regulatory and operational monitoring tasks. This study evaluates three autoencoder architectures — Vanilla Autoencoder (AE), Convolutional Autoencoder (CAE), and Variational Autoencoder (VAE) — for anomaly detection using waterfall image representations derived from FM-band (86.5–108 MHz) measurement data. The models were assessed both as standalone reconstruction-based detectors and as feature extractors combined with supervised classifiers. Standalone autoencoders achieved F1-scores between 0.794 and 0.813. Higher performance was obtained when encoder-derived latent representations were used with supervised models, where the best result (F1-score: 0.915) was achieved by the CAE + ExtraTrees combination. The results indicate that hybrid encoder–classifier approaches can provide an effective practical solution for anomaly detection in spectrum monitoring environments. While promising, the findings are limited to the investigated FM-band dataset and require further validation across broader spectrum environments.

DOI: 10.36244/ICJ.2026.2.7
Download 

Raja Ouadad, and Hicham Mouncif
MTL-BERTNet: A Multi-Task Learning Framework for Aspect and Sentiment Analysis in MOOC Reviews 

In the context of large-scale online learning environments, analyzing student feedback is crucial for improving course content and learner engagement. This paper proposes MTL-BERTNet, a novel multi-task learning architecture that jointly performs aspect category classification and sentiment polarity detection from MOOC reviews. The model leverages contextual embeddings from a pre-trained BERT encoder and integrates a convolutional multi-head attention mechanism to capture subtle semantic nuances and inter-task dependencies. To further enhance shared representation learning across tasks, an inter-task matching layer (IML) is introduced. Experiments conducted on an imbalanced MOOC review dataset demonstrate strong performance, with macro F1-scores of 0.90 for aspect classification and 0.93 for sentiment prediction. These results highlight the effectiveness of jointly modeling aspects and sentiment, offering practical insights for improving course design, instructional quality, and learner satisfaction in MOOC platforms.

DOI: 10.36244/ICJ.2026.2.8
Download 

Rexhep Mustafovski, Tomislav Shuminoski, and Aleksandar Risteski
YOLOv8-Based Firearm Detection for Real-Time Video Surveillance Application 

Real-time firearms detection is key for ensuring safety in public spaces and security-sensitive environments. In this paper, we present a computer vision approach using the YOLOv8 object detection model to identify whether an individual is carrying a firearm. The model’s architecture allows for rapid and accurate detection, making it suitable for real-time applications. A custom dataset of images was used for training and validation, with preprocessing techniques applied to enhance model generalization. The proposed system was evaluated on key performance metrics, achieving high precision and recall in detecting both exposed and partially concealed firearms. The ability to operate in real-time offers a promising solution for improving security measures in various contexts. Future work will focus on refining the system to reduce false positives and expanding the dataset to include a broader range of scenarios.

DOI: 10.36244/ICJ.2026.2.9
Download 

Vincent Karovič, Olena Semenova, Maksym Prytula, Volodymyr Martyniuk, and Vladyslav Kuzniak
Optimized Secure Clustering Scheme for Wireless Sensor Networks 

Wireless Sensor Networks (WSNs) are increasingly employed in various applications, that require efficient, energy-saving and secure transmitting mechanisms. Clustering is a key technique in WSN to enhance scalability, reduce energy consumption, and manage communication. However, tradi-tional clustering algorithms often lack adaptability to alterable wireless environments. This study presents an optimized secure clustering scheme for Wireless Sensor Networks that employs a fuzzy inference system to perform adaptive and secure cluster head selection. The proposed scheme is designed for WSN with a static topology, where sensor nodes remain fixed after de-ployment. This assumption matches Internet of Medical Things (IoMT) applications, such as patient monitoring systems and smart hospital infrastructure, where sensors operate in fixed locations. The proposed fuzzy inference system evaluates parameters of sensor nodes to select optimal cluster heads while mitigating malicious node participation. Next, a genetic algorithm was applied to optimize the parameters of the fuzzy inference system, including the membership function shapes and rule base indices, in order to enhance the accuracy of the proposed clustering scheme. MATLAB simulations demonstrate that the proposed scheme effectively identifies compromised sensor nodes and prevents them from assuming the cluster head role. The security validation confirms that the scheme achieves a high detection rate against internal attacks while extending overall network lifetime and maintaining the low latency. This study provides a feasible solution for secure and intelligent clustering in resource- constrained WSN.

DOI: 10.36244/ICJ.2026.2.10
Download 

Attila Frankó, Gergely Hollósi, Dániel Ficzere, and Pál Varga
Synchronized Real-time Communication In The Eclipse Arrowhead Framework 

Real-time, deterministic communication is a common requirement in most time-critical industrial applications. Due to the increasing complexity, interoperability of such systems must be handled by using appropriate design principles and convenient tools that enhance and ease composability. The Eclipse Arrowhead Framework aims to be a basis for such industrial Cyber-Physical System of Systems by providing a standard for services and enabling interoperability between systems and services. However, precise timing requirements cannot be handled on a service basis as they are closely related to hardware capabilities and the structure of the physical underlying network. This paper addresses this by proposing an extension of the Arrowhead framework to bridge synchronization between instances to the service level without losing any abstraction, thus preserving fundamental achievements of the Arrowhead, such as late binding and loose coupling.

DOI: 10.36244/ICJ.2026.2.11
Download