Automatic modulation classification (AMC) is essential for spectrum awareness, cognitive radio, and electronic intelligence. While classical likelihood- and feature-based methods degrade under channel impairments, deep learning models, though powerful, often ignore domain knowledge and treat frames independently. We propose a Domain Aware Depthwise Separable Temporal Convolutional Network with Multi-Head Attention (DSTCN–MHA) that integrates statistical signal features with learned temporal embeddings and aggregates multiple frames to emphasize informative segments. Two variants are introduced: a light model optimized for efficiency and a heavy model targeting accuracy. Experiments on the RML22.01A dataset show that even at K=1 frames, both models outperform CNN and ResNet baselines, while at K=6 they achieve over 20% gains at low signal-to-noise ratio (SNR) and near-perfect accuracy above 10 dB, with significantly fewer parameters than prior models.

Autonomous vehicles (AVs) promise substantial gains in safety, reliability, and decarbonization, yet safe and efficient interaction in dynamic, heterogeneous traffic remains a key barrier to large-scale deployment. Deep reinforcement learning (DRL) has emerged as a data-driven approach for learning adaptive decision policies that handle complex, unpredictable environments better than rule-based methods. However, different scenarios impose distinct requirements, necessitating scenario-specific algorithms. This survey systematically reviews DRL for four typical scenarios (highways, on-ramp merging, roundabouts, and unsignalized intersections), summarizes road features and recent advances, and evaluates methods using five criteria: driving safety, driving efficiency, training efficiency, unselfishness, and interpretability (DDTUI). Each DDTUI criterion is analyzed with respect to the reviewed algorithms. In addition, a dedicated scenario-centric learning transferability analysis is introduced that systematically evaluates whether each reviewed method demonstrates scene-specific learning improvements and assesses how effectively their designs transfer across the four scenarios. Finally, the challenges for future DRL-based decision-making algorithms are summarized.

The Open Radio Access Network (O-RAN) architecture offers an innovative approach to wireless network design by enabling multi-vendor compatibility and dynamic resource allocation. However, extensive network configurations and high data traffic volumes present significant challenges to its sustainability in terms of energy consumption. Therefore, in this study, we develop a 5G O-RAN traffic steering and sleep mode control system based on Deep Q-Learning (DQN), targeting low-traffic scenarios where significant energy savings can be achieved through selective activation and deactivation of devices. The proposed scheme leverages a deep reinforcement learning algorithm to optimize the mappings from User Equipments (UEs) to Radio Units (RUs), from RUs to Distributed Units (DUs), and from DUs to Centralized Units (CUs), in order to maximize energy savings in O-RAN. Using UE Reference Signal Received Power (RSRP) and RU load levels as input states, the system generates energy-efficient mapping actions for dynamic traffic steering and sleep mode control. The simulation results demonstrate that the proposed mapping methods achieve 5∼12% energy savings compared to the benchmark scenario.

State-Of-the-art deep transfer learning methods depend on exhaustive, trial-and-error fine-tuning of pre-trained models—a process that is both computationally expensive and unreliable when data in target domain are scarce. To overcome these limitations, we propose a domain-informed fine-tuning strategy built upon a novel Innately-Intelligent Neural Network (IINN) architecture. Unlike how state-of-the-art deep learning models are heuristically constructed, IINN constructs each layer in a domain informed manner by directly mapping the mathematical operations of analytical equations (e.g., 3GPP propagation models) into it’s network architecture prior to any training. This "innate" design strategy inherently aligns each layer with specific physical parameters, making the model fully interpretable. As a result, we can pre-identify the exact layers associated with parameters that change between source and target domains and fine-tune only those—eliminating the need for iterative layer-by-layer retraining. This targeted fine tuning approach reduces computational overhead and data requirements. We validated IINN on radio-propagation modelling for cellular networks, achieving faster adaptation and higher accuracy than the conventional fine-tuning approach. Experimental evaluations demonstrate that our proposed domain-aware transfer learning framework achieves up to 16.4% improvement in sector-based performance and approximately 10.3% gain in adapting to varying base station heights, with overall average gains in the 10–15% range over state-of-the-art DNN transfer learning approaches. The proposed framework offers a promising direction for data-efficient learning in next-generation wireless systems.

The General Data Protection Regulation (GDPR), the EU/UK data protection legislation, has necessitated a critical need for compliance modelling to meet its strict and sophisticated requirements. Traditional techniques for modelling security and privacy-related threats fall short of addressing and mitigating the threats of non-compliance. This paper introduces a practical solution to modelling GDPR-compliance based on Defeasible Logic Programming (DeLP), which enhances the robustness and reasoning capabilities of compliance models in real-world scenarios. Furthermore, to overcome the challenges of UNDECIDED query outputs in logical reasoning, we incorporate explicit priorities for conflicting rules and suggest related knowledge for a query in an incomplete knowledge base. To finalize the compliance modelling system, we develop the threat mitigation mechanism that specifies the reasons in case there is a non-compliance threat, along with the suggested actions to mitigate the threats. The application of our approach is demonstrated through a case study on Fitbit , health tracking devices, focusing on non-compliance threats and resolving ”UNDECIDED” query results. Our findings show that the inference engine efficiently identifies non-compliance threats, handles UNDECIDED query results, and suggests appropriate threat mitigation measures. • Developed a knowledge base using Defeasible Logic Programming (DeLP) for GDPR compliance. • Integrated a DeLP-based reasoning mechanism to identify and mitigate non-compliance threats. • Implemented handlers for resolving contradictions in the knowledge base. • Integrated mechanisms to address incompleteness in the knowledge base. • Validated the approach with a Fitbit case study addressing non-compliance threats.

Intelligent transportation systems increasingly depend on wireless communication for broadcasting traffic messages and facilitating real-time vehicular communication. In this context, message authentication is crucial for establishing secure and reliable communication. However, security solutions must consider the dynamic nature of vehicular communication links, which fluctuate between line-of-sight (LoS) and non-line-of-sight (NLoS) due to obstructions. This paper proposes a lightweight cross-layer authentication scheme that employs public-key infrastructure (PKI)-based authentication for initial legitimacy detection/handshaking while using key-based physical-layer re-authentication for message verification. This approach reduces signature generation and signaling overheads associated with each transmission, thereby enhancing network scalability. However, the receiver operating characteristic (ROC;Pd: detection vs.PFA: false alarm probabilities) of the latter decreases with lower signal-to-noise ratio (SNR). To address this, we investigate the use of reconfigurable intelligent surfaces (RISs) to strengthen the SNR directed toward the designated vehicle in shadowed areas (i.e., NLoS scenarios), thereby improving the ROC. Theoretical analysis and practical implementation are conducted using a 1-bit RIS consisting of 64×64 reflective metasurfaces. Experimental results show a significant improvement inPd, increasing from 0.82 to 0.96 at SNR = −6 dB for an orthogonal frequency-division multiplexing (OFDM) system with 128 subcarriers. We also conducted informal and formal security analyses using Burrows-Abadi-Needham (BAN) logic to prove the scheme’s ability to resist passive and active attacks. Furthermore, the proposed scheme reduces computational and communication overheads by 43% and 13%, respectively, compared to traditional cryptographic methods, demonstrating its superiority for real-time, challenging communication scenarios.

The General Data Protection Regulation (GDPR), an EU data protection law, requires compliance modeling techniques to help service providers meet its stringent requirements. Traditional privacy modeling techniques often fail to address and mitigate threats of non-compliance. This paper introduces an efficient threat modeling technique based on Defeasible Logic Programming (DeLP) to identify and mitigate non-compliance threats. To achieve this, we construct a DeLP-based knowledge base that integrates facts and rules derived from GDPR requirements. We then implement an inference engine to reason about GDPR non-compliance threats upon this knowledge base. Two novel concepts, namely the horizontal complexity and vertical complexity of a DeLP knowledge base, have been defined to further analyze and evaluate the complexity of the proposed DeLP-based modeling mechanism. An empirical demonstration validates the system’s feasibility and confirms the time complexity of the proposed reasoner. The findings demonstrate that the proposed DeLP-based technique provides an effective approach to GDPR compliance modeling and improves legal reasoning.

In the rapidly evolving landscape of wireless networks, accurate and resilient propagation models are essential to achieve optimal performance and reliability. This paper presents a novel domain-aware framework for interpretable and resilient propagation models. The proposed approach represents an innovative architecture framework that is not only interpretable but can also deal with training data size scarcity. Bridges domain knowledge with machine learning. The proposed approach leverages a combination of domain expertise, analytical modeling, and customized neural networks to construct interpretable models that excel in both identical distribution and non-identical distribution test-train dataset scenarios. Through a comprehensive analysis, we demonstrate the proposed approach's ability to adapt and refine models in response to real-world variations, ensuring consistent, high-quality performance. The proposed framework not only enhances our understanding of complex systems but also paves the way for the creation of digital twins for wireless networks. Furthermore, the root mean square error of the performance metric for the proposed approach is reported as 6.97 dB, further confirming its effectiveness in accurately predicting the results of wireless propagation.

Satellite-assisted internet of things (IoT) networks have emerged as a beacon of promise, offering global coverage and uninterrupted connectivity. However, the challenges of resource allocation and task offloading in such networks are intricate due to the unique characteristics of satellite communication systems. This research’s findings enrich the landscape of energy-efficient and dependable satellite-assisted IoT networks. The paper navigates the delicate balance between energy efficiency, network throughput, and fairness in distributing resources among IoT devices. The proposed techniques, notably the Outer Approximation Algorithm (OAA), usher in seamless connectivity and resource optimization. The central challenge at hand, a concave fractional programming problem, transforms through the Charnes–Cooper transformation, presenting as a concave optimization enigma. Herein, the proposed outer approximation algorithm takes flight, navigating the intricate paths of concave optimization. The performance of the epsilon-optimal solution faces scrutiny under diverse system parameters—the constellation of IoT devices, their affiliations, fairness considerations, and the equitable distribution of resource blocks. This contribution not only enriches research but also opens doors to the boundless possibilities of satellite-assisted IoT networks.

Data-driven applications and services have been increasingly deployed in all aspects of life including healthcare and medical services in which a huge amount of personal data is collected, aggregated, and processed in a centralised server from various sources. As a consequence, preserving the data privacy and security of these applications is of paramount importance. Since May 2018, the new data protection legislation in the EU/UK, namely the General Data Protection Regulation (GDPR), has come into force and this has called for a critical need for modelling compliance with the GDPR's sophisticated requirements. Existing threat modelling techniques are not designed to model GDPR compliance, particularly in a complex system where personal data is collected, processed, manipulated, and shared with third parties. In this paper, we present a novel comprehensive solution for developing a threat modelling technique to address threats of non-compliance and mitigate them by taking GDPR requirements as the baseline and combining them with the existing security and privacy modelling techniques (i.e., STRIDE and LINDDUN, respectively). For this purpose, we propose a new data flow diagram integrated with the GDPR principles, develop a knowledge base for the non-compliance threats, and leverage an inference engine for reasoning the GDPR non-compliance threats over the knowledge base. Finally, we demonstrate our solution for threats of non-compliance with legal basis and accountability in a telehealth system to show the feasibility and effectiveness of the proposed solution.

Intelligent transportation systems require a reliable exchange of information between network terminals in different vehicular communication environments. Making effective use of the dedicated spectrum is crucial to maximizing communication performance. This requires optimising the modulation order according to different channel conditions. This paper proposes a lightweight spectrum awareness methodology that uses wideband spectrum monitoring and deep learning-based modulation classification techniques to optimise the modulation order. We introduce a channel quality indicator block in which the classifier’s accuracy of detection is used as a forward indicator for the choice of the best modulation type for transmission. By using a 3D stochastic vehicular channel, we evaluate the classification performance at different channel parameter settings, including, speed, variance, and signal-to-noise ratio in urban and rural areas. The experimental analyses demonstrate the capability of the proposed approach to supporting a high detection probability for acceptable false decision-making ≤ 20%.

Intelligent transportation systems are an emerging technology that facilitates real-time vehicle-to-everything communication. Hence, securing and authenticating data packets for intra- and inter-vehicle communication are fundamental security services in vehicular ad-hoc networks (VANETs). However, public-key cryptography (PKC) is commonly used in signature-based authentication, which consumes significant computation resources and communication bandwidth for signatures generation and verification, and key distribution. Therefore, physical layer-based secret key extraction has emerged as an effective candidate for key agreement, exploiting the randomness and reciprocity features of wireless channels. However, the imperfect channel reciprocity generates discrepancies in the extracted key, and existing reconciliation algorithms suffer from significant communication costs and security issues. In this paper, PKC-based authentication is used for initial legitimacy detection and exchanging authenticated probing packets. Accordingly, we propose a blockchain-based reconciliation technique that allows the trusted third party (TTP) to publish the correction sequence of the mismatched bits through a transaction using a smart contract. The smart contract functions enable the TTP to map the transaction address to vehicle-related information and allow vehicles to obtain the transaction contents securely. The obtained shared key is then used for symmetric key cryptography (SKC)-based authentication for subsequent transmissions, saving significant computation and communication costs. The correctness and security robustness of the scheme are proved using Burrows–Abadi–Needham (BAN)-logic and Automated Validation of Internet Security Protocols and Applications (AVISPA) simulator. We also discussed the scheme’s resistance to typical attacks. The scheme’s performance in terms of packet delay and loss ratio is evaluated using the network simulator (OMNeT++). Finally, the computation analysis shows that the scheme saves ∼99% of the time required to verify 1000 messages compared to existing PKC-based schemes.

Artificial Intelligence-based applications have been increasingly deployed in every field of life including smart homes, smart cities, healthcare services, and autonomous systems where personal data is collected across heterogeneous sources and processed using ”black-box” algorithms in opaque centralised servers. As a consequence, preserving the data privacy and security of these applications is of utmost importance. In this respect, a modelling technique for identifying potential data privacy threats and specifying countermeasures to mitigate the related vulnerabilities in such AI-based systems plays a significant role in preserving and securing personal data. Various threat modelling techniques have been proposed such as STRIDE, LINDDUN, and PASTA but none of them is sufficient to model the data privacy threats in autonomous systems. Furthermore, they are not designed to model compliance with data protection legislation like the EU/UK General Data Protection Regulation (GDPR), which is fundamental to protecting data owners’ privacy as well as to preventing personal data from potential privacy-related attacks. In this article, we survey the existing threat modelling techniques for data privacy threats in autonomous systems and then analyse such techniques from the viewpoint of GDPR compliance. Following the analysis, We employ STRIDE and LINDDUN in autonomous cars, a specific use-case of autonomous systems, to scrutinise the challenges and gaps of the existing techniques when modelling data privacy threats. Prospective research directions for refining data privacy threats & GDPR-compliance modelling techniques for autonomous systems are also presented.

In recent years, research has focused on exploiting the inherent physical (PHY) characteristics of wireless channels to discriminate between different spatially separated network terminals, mitigating the significant costs of signature-based techniques. In this paper, the legitimacy of the corresponding terminal is firstly verified at the protocol stack’s upper layers, and then the re-authentication process is performed at the PHY-layer. In the latter, a unique PHY-layer signature is created for each transmission based on the spatially and temporally correlated channel attributes within the coherence time interval. As part of the verification process, the PHY-layer signature can be used as a message authentication code to prove the packet’s authenticity. Extensive simulation has shown the capability of the proposed scheme to support high detection probability at small signal-to-noise ratios. In addition, security evaluation is conducted against passive and active attacks. Computation and communication comparisons are performed to demonstrate that the proposed scheme provides superior performance compared to conventional cryptographic approaches.

Heterogeneity is a key feature of 5G and beyond networks for Internet of things applications in various fields. Regarded as the leading low power wide area network, long range wide area network (LoRaWAN) is expected to accomplish 5G's massive machine-type communications target by integrating it into 5G network. In this paper, we design and implement a LoRaWAN-5G integrated network with a collaborative Radio Access Network and a converged core network. We built a 5G-based LoRaWAN gateway that communicates with 5G new radio. To the best of our knowledge, this is the first LoRaWAN gateway that uses 5G network as its backhaul. Moreover, the LoRaWAN servers are deployed within the core network of the 5G testbed, enhancing the security and privacy of LoRaWAN data. This hybrid network has been deployed to monitor the heating system of rooms in James Watt South Building at the University of Glasgow, demonstrating the stability, high flexibility and low deployment cost of the network.

The increasing demand for throughput, ultra-low latency, ultra-high reliability, and ubiquitous coverage have made researchers explore several novel solutions to set the basis for future generations of wireless communications. These demands, however, will consume a significant amount of resources, particularly in the case of cell-free millimetre-wave (mm-Wave) massive multiple input multiple output systems (MIMO), which is the promising approach for future wireless generations. In this paper, we propose a novel and low-complexity matching approach to dynamically activate a set of radio frequency (RF) chains based on the Hungarian algorithm to maximize the total energy efficiency in the uplink of the cell-free mm-Wave massive MIMO systems. Simulation results demonstrate that our proposed scheme achieves up to 13.5%, 20% and 58.7% energy efficiency improvement compared to state-of-the-art adaptive RF chains activation (ARFA), random access point activation and fixed activation scheme when all RF chains at each AP are switched on, respectively. In addition, compared to the ARFA scheme, the proposed matching scheme achieves a complexity reduction ratio of up to 189.6%.

In a LoRaWAN network one of the main reasons of packet outage is the destructive interference that is caused by colliding packets. As the network operates with an ALOHA-like channel access setup, there is no easy way of preventing two or more devices transmitting at the same time, possibly generating interference to each other. Different methods are proposed in literature that can be used to decrease this chance. However, most of them require extensive use of downlink messages coupled with involved algorithms at the network side, often for only a marginal improvement in performance. In this paper we analyse some ways to optimise the Packet Delivery Ratio (PDR) of a LoRaWAN network that can be used when setting up a node or a group of nodes, do not involve downlink and can operate without knowledge of other devices in the same network. These are shown to provide a small boost in performance of maximum 10%, which is akin to that of more complex, downlink-dependant schemes, while decreasing the set up complexity considerably.

Narrow Band Internet of Things (NB-IoT) is one of the drivers of industry 4.0 and the study of its wireless behavior in Indoor industrial environments has become important. This is because of the unfavorable conditions poised to wireless propagation as a result of the presence of heavy-duty equipment and the physical structure of industrial buildings. In this paper, an indoor industry was modeled to present some of the reflective characteristics and its effects on wireless propagation, particularly large-scale fading. The results obtained which include propagation paths, path loss, and impulse response showed how the environment affected the wireless transmission of NB-IoT. However, to mitigate this challenge, a collaborative scheme is introduced to improve the transmission among the affected NB-IoT terminals. The proposed scheme resulted in a collective improvement in path loss value by 30.44%.

Securing data exchange between intercommunicating terminals, e.g., vehicle-to-everything, constitutes a technological challenge that needs to be addressed. Security solutions must be computationally efficient and flexible enough to be implemented in any wireless propagation environment. Recently, physical layer security has gained popularity, which exploits the randomness of wireless channel responses for extracting high entropy secret cryptographic keys. The current state-of-the-art relies on the independently varying channel sources of randomness, e.g., received signal strength (RSS) and phase. However, the limited capability of RSS-based extraction techniques has motivated researchers to investigate alternative approaches. Although phase-based approaches have emerged in many studies, optimising the extraction performance by adapting the algorithm to the non-reciprocal components of static and dynamic channels remains a challenge. In this paper, we propose an adaptive multilevel quantisation approach that adjusts the size of the quantisation region to the channel responses’ non-reciprocity parameters, thus optimising the trade-off between the bit generation rate (BGR) and the bit mismatch rate (BMR). The probability of error has been theoretically formulated. Accordingly, the order of the quantisation process is adapted for acceptable mismatching probability. Moreover, simulation analysis is conducted to prove the ability of the proposed approach to provide flexible adaptation of the quantisation order at different signal-to-noise ratios (SNRs), achieving fast secret bit generation rates 1. 1$\sim$2.85bits/packet at SNRs of 10$\sim$25 dB for acceptable BMR $\leq 0.1$.

In real-time wireless control systems, the freshness of information is crucial since performance highly depends on timely exchange of information between the plant and the controller. In this study, the age metric, Age of Control Process (AoCP), is proposed for real-time wireless control systems which is different from traditional Age of Information (AoI). The First Generate First Serve (FGFS) M/M/1/1 → M/M/1/1 tandem queue model is considered to present a closed form expression for computation of AoCP and testbed environment is employed for the analysis of AoCP measure. Our experiment results show that FGFS M/M/1/1 → M/M/1/1 queuing model is suitable to represent real-time control systems. In addition, the proposed definition provides closer results to the experiment results when compared with the traditional AoI definition. To the best of our knowledge, this is the first study that conducts information freshness measurements on real-time control testbed.

Millimeter-wave (mm-wave) communication, which has already been a part of the fifth generation of mobile communication networks (5G), would result in ultra dense small cell deployments due to its limited coverage characteristics. To enable seamless handovers between indoor and outdoor environments, a mobility prediction of an indoor user is studied by deploying Markov chains. Based on the effect of external factors on the user's mobility, a simulation scenario is created to model the trajectory of an indoor user w.r.t the most visited areas before leaving the indoor environment. Based on that, a method for initializing the transition matrix of Markov chains is proposed, via Q-learning. The proposed solution is compared to a standard online learning Markov chain model in terms of different mobility models and learning rates. Results show that the proposed solution is always able to outperform the standard method in terms of prediction accuracy.

Day by day innovation in wireless communications and micro-technology has evolved in the development of wireless sensor networks. This technology has applications such as healthcare supervision, home security, battlefield surveillance and many more. However, due to the use of small batteries with low power this technology faces the issue of power and target monitoring. There is much research done to overcome these issues with the development of different architecture and algorithms. In this paper, a scheduling machine learning algorithm called adaptive learning automata algorithm(ALAA) is used. It provides an efficient scheduling technique. Such that each sensor node in the network has been equipped with learning automata, and with this, they can select their proper state at any given time. The state of the sensor is either active or sleep. For the experiment, different parameters are used to check the consistency of the algorithm to schedule the sensor node such that it can cover all the targets with the use of less power. The results obtained from the experiments show that the proposed algorithm is an efficient way to schedule the sensor nodes to monitor all the targets with use of less power. On the whole, this paper manages to achieve its goal by contributing to the related research on wireless sensor networks with a new design of a learning automata scheduling algorithm. The ability of this proposed algorithm to use the minimum number of sensors to be in active state verified to reduce the use of power in the network. Thus, achieving the goal by enhancing the lifetime of wireless sensor networks.

Vehicular safety applications have much significance in preventing road accidents and fatalities. Among others, cellular networks have been under investigation for the procurement of these applications subject to stringent requirements for latency, transmission parameters, and successful delivery of messages. Earlier contributions have studied utilization of Long‐Term Evolution (LTE) under single cell, Friis radio, or simplified higher layer. In this paper, we study the utilization of LTE under multicell and multipath fading environment and introduce the use of adaptive awareness range. Then, we propose an algorithm that uses the concept of quality of service (QoS) class identifiers (QCIs) along with dynamic adaptive awareness range. Furthermore, we investigate the impact of background traffic on the proposed algorithm. Finally, we utilize medium access control (MAC) layer elements in order to fulfill vehicular application requirements through extensive system‐level simulations. The results show that, by using an awareness range of up to 250 m, the LTE system is capable of fulfilling the safety application requirements for up to 10 beacons/s with 150 vehicles in an area of 2 × 2 km2. The urban vehicular radio environment has a significant impact and decreases the probability for end‐to‐end delay to be ≤100 ms from 93%–97% to 76%–78% compared to the Friis radio environment. The proposed algorithm reduces the amount of vehicular application traffic from 21 Mbps to 13 Mbps, while improving the probability of end‐to‐end delay being ≤100 ms by 20%. Lastly, use of MAC layer control elements brings the processing of messages towards the edge of network increasing capacity of the system by about 50%.