Although multi-tier vehicular Metaverse promises to transform vehicles into essential nodes -- within an interconnected digital ecosystem -- using efficient resource allocation and seamless vehicular twin (VT) migration, this can hardly be achieved by the existing techniques operating in a highly dynamic vehicular environment, since they can hardly balance multi-objective optimization problems such as latency reduction, resource utilization, and user experience (UX). To address these challenges, we introduce a novel multi-tier resource allocation and VT migration framework that integrates Graph Convolutional Networks (GCNs), a hierarchical Stackelberg game-based incentive mechanism, and Multi-Agent Deep Reinforcement Learning (MADRL). The GCN-based model captures both spatial and temporal dependencies within the vehicular network; the Stackelberg game-based incentive mechanism fosters cooperation between vehicles and infrastructure; and the MADRL algorithm jointly optimizes resource allocation and VT migration in real time. By modeling this dynamic and multi-tier vehicular Metaverse as a Markov Decision Process (MDP), we develop a MADRL-based algorithm dubbed the Multi-Objective Multi-Agent Deep Deterministic Policy Gradient (MO-MADDPG), which can effectively balances the various conflicting objectives. Extensive simulations validate the effectiveness of this algorithm that is demonstrated to enhance scalability, reliability, and efficiency while considerably improving latency, resource utilization, migration cost, and overall UX by 12.8%, 9.7%, 14.2%, and 16.1%, respectively.

Metaverse has generated significant interest for enabling wireless systems due to its self-sustainability and proactive analytic capabilities. It enables the deployment of meta spaces featuring avatars and digital twins for various real-world applications. However, it has become challenging to efficiently deploy meta spaces on edge/cloud environments while managing communication and computing resources effectively. In this paper, we propose a novel network virtualization framework within a shared system to make the deployment of meta spaces for various metaverse applications cost-efficient. The framework abstracts, isolates, and facilitates the sharing of wireless and computing resources, thereby enhancing flexibility and efficiency for metaverse-driven applications. It involves three key players: network operators (selling resources), metaverse operators (managing transactions), and end-users (purchasing resources). An optimization problem is formulated to optimize the wireless resource allocation, computing resource allocation, association of end-devices with meta spaces deployed at the network operators' base stations, communication resource cost, and computing resource cost. We address the formulated NP-Hard mixed integer non-linear programming (MINLP) problem using decomposition, convex optimization, and hierarchical matching. Numerical results confirm the validity of the proposed approach.

Determining the ideal architecture for deep learning models, such as the number of layers and neurons, is a difficult and resource-intensive process that frequently relies on human tuning or computationally costly optimization approaches. While Particle Swarm Optimization (PSO) and Large Language Models (LLMs) have been individually applied in optimization and deep learning, their combined use for enhancing convergence in numerical optimization tasks remains underexplored. Our work addresses this gap by integrating LLMs into PSO to reduce model evaluations and improve convergence for deep learning hyperparameter tuning. The proposed LLM-enhanced PSO method addresses the difficulties of efficiency and convergence by using LLMs (particularly ChatGPT-3.5 and Llama3) to improve PSO performance, allowing for faster achievement of target objectives. Our method speeds up search space exploration by substituting underperforming particle placements with best suggestions offered by LLMs. Comprehensive experiments across three scenarios—(1) optimizing the Rastrigin function, (2) using Long Short-Term Memory (LSTM) networks for time series regression, and (3) using Convolutional Neural Networks (CNNs) for material classification—show that the method significantly improves convergence rates and lowers computational costs. Depending on the application, computational complexity is lowered by 20% to 60% compared to traditional PSO methods. Llama3 achieved a 20% to 40% reduction in model calls for regression tasks, whereas ChatGPT-3.5 reduced model calls by 60% for both regression and classification tasks, all while preserving accuracy and error rates. This groundbreaking methodology offers a very efficient and effective solution for optimizing deep learning models, leading to substantial computational performance improvements across a wide range of applications.

In response to the success of proprietary Large Language Models (LLMs) such as OpenAI's GPT-4, there is a growing interest in developing open, non-proprietary LLMs and AI foundation models (AIFMs) for transparent use in academic, scientific, and non-commercial applications. Despite their inability to match the refined functionalities of their proprietary counterparts, open models hold immense potential to revolutionize healthcare applications. In this paper, we examine the prospects of open-source LLMs and AIFMs for developing healthcare applications and make two key contributions. Firstly, we present a comprehensive survey of the current state-of-the-art open-source healthcare LLMs and AIFMs and introduce a taxonomy of these open AIFMs, categorizing their utility across various healthcare tasks. Secondly, to evaluate the general-purpose applications of open LLMs in healthcare, we present a case study on personalized prescriptions. This task is particularly significant due to its critical role in delivering tailored, patient-specific medications that can greatly improve treatment outcomes. In addition, we compare the performance of open-source models with proprietary models in settings with and without Retrieval-Augmented Generation (RAG). Our findings suggest that, although less refined, open LLMs can achieve performance comparable to proprietary models when paired with grounding techniques such as RAG. Furthermore, to highlight the clinical significance of LLMs-empowered personalized prescriptions, we perform subjective assessment through an expert clinician. We also elaborate on ethical considerations and potential risks associated with the misuse of powerful LLMs and AIFMs, highlighting the need for a cautious and responsible implementation in healthcare.

The advancements in autonomous driving technology, coupled with the growing interest from automotive manufacturers and tech companies, suggest a rising adoption of Connected Autonomous Vehicles (CAVs) in the near future. Despite some evidence of higher accident rates in AVs, these incidents tend to result in less severe injuries compared to traditional vehicles due to cooperative safety measures. However, the increased complexity of CAV systems exposes them to significant security vulnerabilities, potentially compromising their performance and communication integrity. This paper contributes by presenting a detailed analysis of existing security frameworks and protocols, focusing on intra- and inter-vehicle communications. We systematically evaluate the effectiveness of these frameworks in addressing known vulnerabilities and propose a set of best practices for enhancing CAV communication security. The paper also provides a comprehensive taxonomy of attack vectors in CAV ecosystems and suggests future research directions for designing more robust security mechanisms. Our key contributions include the development of a new classification system for CAV security threats, the proposal of practical security protocols, and the introduction of use cases that demonstrate how these protocols can be integrated into real-world CAV applications. These insights are crucial for advancing secure CAV adoption and ensuring the safe integration of autonomous vehicles into intelligent transportation systems.

In the dynamic vehicular metaverse, delivering a seamless user experience (UX) and effective human-machine interaction (HMI) is challenging due to vehicle mobility and varying resource needs. This paper introduces an adaptive resource allocation and twin migration framework using Multi-Agent Deep Reinforcement Learning (MADRL) for a multi-tier vehicular metaverse. The framework enables cooperative agents to dynamically allocate resources and migrate vehicle twins across vehicle, edge, and cloud layers, ensuring seamless UX and efficient HMI. The joint resource allocation and twin migration optimization problem is modeled as MDP and a hierarchical multi-agent deep deterministic policy gradient-with QMIX (MADDPG-Q) strategy is adopted to solve it, reducing latency and optimizing resource use. Moreover, the proposed framework is designed to be context-aware, adjusting HMI based on real-time conditions, and enhancing interaction quality. Simulation results show significant improvements in UX, latency reduction, and resource efficiency.

Due to the advancements in technology, the number of connected devices in a data processing and computing environment of edge and Internet of Things (IoT) has increased exponentially. Thus, proliferation of electronics by billions has raised serious security concerns. Therefore, designing effective coordination mechanisms and ensuring user verification and data security pose significant challenges in edge environments. Moreover, dispersed architecture of fog computing presents challenges related to privacy, secure authentication, and fog device security. Therefore, to address these drawbacks, in this paper, we propose an novel approach to enhance user verification and data security in edge environment. Our proposed scheme leverages the intrinsic properties of the Ethereum blockchain to develop a smart contract-based solution for registering and verifying digital identities securely to mitigate security and privacy concerns associated with traditional authentication techniques. We demonstrate the effectiveness of the proposed scheme via experiments and simulations conducted on the Ethereum blockchain environment. The results indicate that the proposed scheme outperform than other state-of-the-art techniques in terms of transaction gas and execution cost for both identity registration and verification.

Clustered Federated Multi-task Learning (CFL) has emerged as a promising technique to address statistical challenges, particularly with non-independent and identically distributed (non-IID) data across users. However, existing CFL studies entirely rely on the impractical assumption that devices possess access to accurate ground-truth labels. This assumption becomes specifically problematic in hierarchical wireless networks (HWNs), with vast unlabeled data and dual-level model aggregation, not only leading to slowing down convergence speeds and extending processing times but also resulting in increased resource consumption. To this end, we propose Clustered Federated Semi-Supervised Learning (CFSL), a novel framework tailored for more realistic scenarios in HWNs. We leverage specialized models resulting from device clustering and present two prediction model schemes, the best-performing specialized model and the weighted-averaging ensemble model, to correctly label unlabeled, unseen data. For the best-performing specialized model scheme, a specialized model excelling in label prediction for a specific device is assigned to correctly label the unlabeled data, even when the data originates from other environments, while the weighted-averaging ensemble model combines all specialized models into a unified model, capturing more details from broader data distributions across edge networks. The CFSL also introduces two novel prediction time schemes, split-based and stopping-based, for accurately timing the labeling process, alongside two strategic device selection schemes, greedy and round-robin, upon reaching each cluster’s stopping point. Extensive testing validates CFSL’s superiority over existing models in labeling and testing accuracies and resource efficiency, achieving up to 51% energy savings.

With its significant security potential, the quantum internet is poised to revolutionize technologies like cryptography and communications. Although it boasts enhanced security over traditional networks, the quantum internet still encounters unique security challenges essential for safeguarding its Confidentiality, Integrity, and Availability (CIA). This study explores these challenges by analyzing the vulnerabilities and the corresponding mitigation strategies across different layers of the quantum internet, including physical, link, network, and application layers. We assess the severity of potential attacks, evaluate the expected effectiveness of mitigation strategies, and identify vulnerabilities within diverse network configurations, integrating both classical and quantum approaches. Our research highlights the dynamic nature of these security issues and emphasizes the necessity for adaptive security measures. The findings underline the need for ongoing research into the security dimension of the quantum internet to ensure its robustness, encourage its adoption, and maximize its impact on society.

In today’s era of massive information exchange, trust plays a significant role in the Internet of Things (IoT) security and privacy. Trust specifies the confidence of one point or network entity on another in a network model. As IoT devices of smart logistics rely on Internet connectivity, potential cyber-attacks pose a genuine threat. IoT devices have constrained capabilities, and therefore it is very difficult to differentiate between trustworthy and malicious IoT devices. The main contribution of this research is, first, to develop a mathematical model of the Analytical Network Process (ANP) for security and privacy in smart logistics. Second, IoT devices are categorized using trust as an attribute in the security domain of the smart logistics framework. Third, the applicability of ANP as a trust model for malicious IoT device identification is evaluated. And forth, for comparison, the Analytical Hierarchal Process (AHP) is applied to the same trust model for malicious IoT device identification, and results are compared with ANP. The proposed mathematical model of ANP for secured smart logistics is designed using three conceptual facets i.e., goal, criteria, and alternatives. The goal of the trust model is to identify malicious IoT devices using multi-criteria based on three trust parameters, i.e., Entity-Oriented Trust Model (EOTM), Data-Oriented Trust Model (DOTM), and Combined Trust Model (CTM). The alternatives in the trust model are the various candidate IoT devices. The architecture of the smart logistics framework has three modules i.e., internal IoT network, border router, and external network. The IoT devices communicate through the border router with the external world. Therefore, IoT devices are scanned by border router and taken as a trusted node in the secured smart logistics framework. Additionally, for comparison and validation, results obtained from ANP and AHP are evaluated. Results reveal a percentage difference of 16.86

In today’s age of digital interconnectedness, understanding and addressing the nuances of social cybersecurity have become paramount. Unlike its broader counterparts, information security and cybersecurity, which are focused on safeguarding all forms of sensitive data and digital systems, social cybersecurity places its emphasis on the human and social dimensions of cyber threats. This field is uniquely positioned to address issues such as different social cybersecurity attacks like cyberbullying, cybercrime, spam, terrorist activities, and community detection. The significance of detection methods in social cybersecurity is underscored by the need for timely and proactive responses to these threats. In this comprehensive review, we delve into various techniques, attacks, challenges, potential solutions, and trends within the realm of detecting social cybersecurity attacks. Additionally, we explore the potential of readily available public datasets and tools that could expedite research in this vital domain. Our objective is not only to tackle the existing challenges but also to illuminate potential pathways for future exploration. Through this survey, our primary focus is to provide valuable insights into the rapidly evolving landscape of social cybersecurity. By doing so, we aim to assist researchers and practitioners in developing effective prediction models, enhancing defense strategies, and ultimately fostering a safer digital environment.

This paper presents a new data clustering technique aimed at enhancing the performance of the trainable path-cost algorithm and reducing the computational complexity of data clustering models. The proposed method facilitates the discovery of natural groupings and behaviours, which is crucial for effective coordination in complex environments. It identifies natural groupings within a set of features and detects the best clusters with similar behaviour in the data, overcoming the limitations of traditional state-of-the-art methods. The algorithm utilises a density peak clustering method to determine cluster centers and then extracts features from paths passing through these peak points (centers). These features are used to train the support vector machine (SVM) to predict the labels of other points. The proposed algorithm is enhanced using two key concepts: first, it employs Q-Generalised Extreme Value (Q-GEV) under power normalisation instead of traditional generalised extreme value distributions, thereby increasing modelling flexibility; second, it utilises the random vector functional link (RVFL) network rather than the SVM, which helps avoid overfitting and improves label prediction accuracy. The effectiveness of the proposed clustering algorithm is evaluated through various experiments, including those on UCI benchmark datasets and real-world data, demonstrating significant improvements across multiple performance metrics, including F1 measure, Jaccard index, purity, and accuracy, highlighting its capability in accurately identifying paths between similar clusters. Its average F1 measure, Jaccard index, purity, and accuracy is measured 76.87%, 56.29%, 80.29%, and 79.64%, respectively.

Managing the growing complexity of Self-Organizing Networks (SONs) in next-generation communication systems requires agile, real-time strategies that can adapt to multidimensional and highly dynamic conditions. Traditional SON management rooted in centralized, rule-based, and static models, struggles to meet these evolving requirements. Recent advances in multi-agent systems (MAS) and Large Language Models (LLMs) enable the design of intelligent and context-aware frameworks that span multiple operational layers. In this paper, we introduce LaMA-SON, an LLM-driven MAS for intelligent SON management. Specifically, LaMA-SON incorporates specialized agents to handle three critical operational functions: traffic management, quality of service (QoS) optimization, and security threat detection. We perform a proof-of-concept evaluation using a real-world network traffic classification dataset, where traffic, security, and QoS optimization agents make decisions based on role-specific features and structured prompts. Our experimental results demonstrate that LaMA-SON achieves higher accuracy and recall while preserving balanced precision-recall trade-offs and outperforms standalone LLM baselines, highlighting the benefits of multi-agent collaboration. Finally, we highlight various open research challenges that require further investigation to fully realize the potential of LLM-based MAS frameworks in network operations management.

KNNOR-Reg is a Python package designed to address the challenge of imbalanced regression. While popular Python packages exist for tackling imbalanced classification, support for imbalanced regression remains limited. Imbalanced regression involves the underrepresentation of important ranges within a continuous target variable. KNNOR-Reg implements an oversampling technique that generates synthetic samples through interpolation between minority class samples and their nearest neighbors. The labels for synthetic samples are computed based on the inverse distance-weighted average of the nearest neighbors’ labels. KNNOR-Reg offers a user-friendly and extensible Python implementation for oversampling imbalanced regression data, aiming to reduce regressor bias and enhance model outcomes.

Deploying transformer models in Personalized Federated Learning (PFL) over wireless networks is challenging due to their large size, which leads to high communication overhead, increased latency, and excessive energy consumption. Traditional pruning and sparsification methods, designed mainly for conventional deep learning architectures, are ineffective for transformers and can cause divergence or degrade performance—especially when applied to self-attention layers or through direct federated averaging. To address these challenges, we propose a novel dual approach called PFL-TPS (PFL with Transformer Pruning and Sparsification). Our approach efficiently reduces communication and computation costs while maintaining model performance, making it suitable for resource-constrained wireless networks. Specifically, we apply adaptive pruning with trainable thresholds to the transformer's Feed-Forward Layers (FFLs), and only these trainable thresholds are shared with the server, resulting in minimal uploaded data. For the Self-Attention Layers (SALs), instead of transmitting bandwidth-intensive model parameters, we employ a server-side hypernetwork that generates personalized parameters based on device-specific embedding vectors sent by the devices, significantly reducing communication overhead and maintaining personalization. Extensive experiments show that PFL-TPS reduces energy consumption by up to 50%, decreases training time by 60.44%, and improves model accuracy by 49.87% compared to baselines in wireless networks.

The terahertz band, spanning from 100 GHz to 10 THz, has garnered significant attention for its potential to support ultra-high transmissions. Moreover, the Internet of Nano Things (IoNT) offers efficient data collection and analysis mechanisms to optimize nanosystems performance for enhance users' experiences and comfort in daily life. In this article, we explore the integration of IoNT and the terahertz band, envisioning ultra-high-speed transmissions in nano-communication systems for improved users' experiences. We introduce the key components of the IoNT networks based on terahertz band and discuss the challenges with respect to data fusion. Further, we propose an efficient data fusion scheme for the hierarchical IoNT networks and evaluate its performance using simulation under channel capacity, noise, path loss, and power control. We further provide comprehensive use cases in which the proposed method can be beneficial. In addition, we discuss potential challenges, limitations, and future research directions related to our proposed solution and IoNT operating over terahertz bands.

Recently, unmanned aerial vehicles (UAVs) have demonstrated exemplary performance in various scenarios, such as search and rescue, smart city services, and disaster response applications. UAVs can facilitate wireless power transfer (WPT), resource offloading, and data collection from ground IoT devices. However, employing UAVs for such applications poses several challenges, including limited flight duration, constrained energy resources, and the age of information of the data collected. To address these challenges, we employ a UAV swarm to maximize energy harvesting (EH) and data rates for IoT devices by optimizing UAV paths and integrating reconfigurable intelligent surfaces (RIS) technology. We tackle critical constraints, including UAV energy consumption, flight duration, and data collection deadlines, by formulating an optimization problem to find optimal UAV paths and RIS phase shifts. Given the complexity of the problem, its combinatorial nature, and the challenges of obtaining an optimal solution through conventional optimization methods, we decompose the problem into two sub-problems, employing deep reinforcement learning (DRL) to optimize EH and particle swarm optimization (PSO) to optimize RIS phase shifts. Our extensive simulations show that the proposed solution outperforms competitive algorithms, including Brute-Force-PSO, AC-PSO, and PPO-PSO algorithms, providing a robust solution for modern IoT applications.

Prior research on intelligent reflection surface (IRS)-assisted unmanned aerial vehicle (UAV) communications has focused on a fixed location for the IRS or mounted on a UAV. The assumption that the IRS is located at a fixed position will prohibit mobile users from maximizing many wireless network benefits, such as data rate and coverage. Furthermore, assuming that the IRS is placed on a UAV is impractical for various reasons, including the IRS’s weight and size and wind speed in severe weather. Unlike previous studies, this study assumes a single UAV and an IRS mounted on a mobile ground vehicle (M-IRS) to be deployed in an Internet-of-Things (IoT) 6G wireless network to maximize the average data rate. Such a methodology for providing wireless coverage using an M-IRS assisted UAV system is expected in smart cities. In this paper, we formulate an optimization problem to find an efficient trajectory for the UAV, an efficient path for the M-IRS, and users’ power allocation coefficients that maximize the average data rate for mobile ground users. Due to its intractability, we propose efficient techniques to help find the optimization problem’s solution. First, we show that our dynamic power allocation technique outperforms the fixed power allocation technique in the network average sum rate. Then we employ the individual movement model (Random Waypoint Model) in order to represent the users’ movements inside the coverage area. Finally, we propose an efficient approach using a Genetic Algorithm (GA) to find an efficient trajectory for the UAV and an efficient path for the M-IRS to provide wireless connectivity for mobile users during their movement. We demonstrate through simulations that our methodology can enhance the average data rate by 15% on average compared with the static IRS and by 25% on average compared to without the IRS system.

In contrast to methods relying on a centralized training, emerging Internet of Things (IoT) applications can employ federated learning (FL) to train a variety of models for performance improvement and improved privacy preservation. FL calls for the distributed training of local models at end-devices, which uses a lot of processing power (i.e., CPU cycles/sec). Most end-devices have computing power limitations, such as IoT temperature sensors. One solution for this problem is split FL. However, split FL has its problems including a single point of failure, issues with fairness, and a poor convergence rate. We provide a novel framework, called hierarchical split FL (HSFL), to overcome these issues. On grouping, our HSFL framework is built. Partial models are constructed within each group at the devices, with the remaining work done at the edge servers. Each group then performs local aggregation at the edge following the computation of local models. End devices are given access to such an edge aggregated model so they can update their models. For each group, a unique edge aggregated HSFL model is produced by this procedure after a set number of rounds. Shared among edge servers, these edge aggregated HSFL models are then aggregated to produce a global model. Additionally, we propose an optimization problem that takes into account the RLA of devices, transmission latency, transmission energy, and edge servers' compute latency in order to reduce the cost of HSFL. The formulated problem is a mixed-integer non-linear programming (MINLP) problem and cannot be solved easily. To tackle this challenge, we perform decomposition of the formulated problem to yield sub-problems. These sub-problems are edge computing resource allocation problem and joint relative local accuracy (RLA) minimization, wireless resource allocation, task offloading, and transmit power allocation sub-problem. Due to the convex nature of edge computing, resource allocation is done so utilizing a convex optimizer, as opposed to a block successive upper-bound minimization (BSUM) based approach for joint relative local accuracy (RLA) minimization, resource allocation, job offloading, and transmit power allocation. Finally, we present the performance evaluation findings for the proposed HSFL scheme.