This paper presents a novel approach to intelligent log management, using large language models (LLMs) to detect and resolve anomalies in real-time. By extending the capabilities of a model like GPT for log analysis, we can enable the automatic detection of anomalies with actionable recommendations. The model was trained on a dataset of system logs with hyperparameters optimized for anomaly detection accuracy. The evaluation shows that the model successfully identified 85% of anomalies with a false positive rate of 5%, significantly outperforming traditional log analysis tools. Further, the model provides actionable recommendations in 90% of cases, which reduce the Mean Time to Recovery (MTTR) by 40%.

This paper extends the recently developed framework of multilinear kernel regression and imputation via manifold learning (MultiL-KRIM) to impute time-varying edge flows in a graph. MultiL-KRIM uses simplicial-complex arguments and Hodge Laplacians to incorporate the graph topology, and exploits manifold-learning arguments to identify latent geometries within features which are modeled as a point-cloud around a smooth manifold embedded in a reproducing kernel Hilbert space (RKHS). Following the concept of tangent spaces to smooth manifolds, linear approximating patches are used to add a collaborative-filtering flavor to the point-cloud approximations. Together with matrix factorizations, MultiL-KRIM effects dimensionality reduction, and enables efficient computations, without any training data or additional information. Numerical tests on realnetwork time-varying edge flows demonstrate noticeable improvements of MultiL-KRIM over several state-of-the-art schemes.

This paper proposes a Physics-Informed Neural Network (PINN) based on the scattering equivalent source method. Through unsupervised learning, it addresses frequency-domain electromagnetic problems on two-dimensional metal boundaries, demonstrating a certain level of generalization. The introduced PINN leverages the equivalent source principle to represent metal boundaries effectively. Additionally, an attention mechanism is employed to establish correlations between different equivalent source sampling points, enabling dynamic input-output relationships.

This work investigates a Holographic Multiple-Input Multiple-Output (HMIMO) communication system consisting of a source and a receiver being two surfaces in the LoS configuration. The closed-form expressions of the electric field and the channel coupling coefficients are obtained leveraging an asymptotic expansion of the Green's function. The derived approximations show satisfactory accuracy in both near-and farfield conditions, across low and high frequencies. These are then employed to obtain insights regarding (i) the spatial distribution of the beams, and (ii) the mapping relation between source and receiver modes while ensuring sufficient orthogonality under prescribed conditions.

Recursive and non-recursive theoretical closed-form expressions of variable step-size incremental least mean square (VSSILMS) algorithm are derived in wireless sensor networks (WSNs). In fact, the theoretical transient and steady-state of the VSS 1 stand 2 nd-moment relations are derived in recursive and non-recursive closed forms adopting the initial design parameters only. Moreover, the tedious task of finding the simulation length can be implied by a quick simulation of the VSS 1s—moment closed-form having the initial design parameters alone in a prior to the real simulation. In the real simulations, the proposed theoretical 1st- and 2nd-moment closed forms are found in a good agreement with the empirical results.

Ransomware is one of the most significant cybersecurity threats to modern networks, leveraging advanced techniques to communicate with Command and Control (C&C) servers to coordinate attacks, exfiltrate data, and issue encryption commands. The novel approach presented in this study provides a comprehensive method for tracing ransomware network traffic to C&C servers, focusing on detecting the specific communication patterns, encryption protocols, and geolocation data that can enhance existing detection systems. The methodology integrates both signature-based and anomaly-based techniques, allowing for the identification of ransomware traffic even when advanced obfuscation methods, such as domain-flux and fastflux techniques, are employed. Through the use of controlled sandbox environments, custom-built C&C server emulators, and sophisticated network traffic analysis, the research outlines key indicators of ransomware communications and provides practical tools for improving network defenses. The results demonstrate how a hybrid approach to detection can significantly increase accuracy while maintaining operational efficiency, offering a tangible framework for mitigating ransomware threats through early identification of C&C traffic. The study also highlights the importance of understanding the geographical distribution of C&C infrastructure and its implications for network security, particularly in regions with lenient cybersecurity regulations. Overall, the findings contribute to a deeper understanding of the ransomware ecosystem and its critical dependence on covert C&C communications, providing a foundation for developing more effective cybersecurity defenses.

This paper unveils the groundbreaking impact of Generative AI (GenAI) as the dawn of a transformative era in 5G/6G networks and beyond. Exploring its disruptive potential across the value chain—from network design to agile and robust automation— we showcase GenAI as a catalyst for innovation and unparalleled efficiency. While tracing its historical journey from conception to practical implementation, the paper positions GenAI not as the sole solution but as the inception of a new era shaping network design, deployment strategies, and synchronous optimization dynamics. We also scrutinize the role of causal AI and transparent frameworks, such as explainable AI and neuro-symbolic AI in fostering trust and seamlessly integrating domain knowledge. Looking ahead beyond GenAI, we envision a future AI landscape composed of semantic communications, collaborative GenAI and discriminative agents. We also examine the challenges related to scalability and complexity that must be overcome to achieve sustainable AI deployments. Finally, we highlight the significance of emerging computing technologies and frameworks, such as quantum and neuromorphic computing, that play a pivotal role in the broader trajectory towards artificial general intelligence.

Due to its flexibility, cloud computing has become essential in modern operational schemes. However, the effective management of cloud resources to ensure cost-effectiveness and maintain high performance presents significant challenges. The pay-as-you-go pricing model, while convenient, can lead to escalated expenses and hinder long-term planning. Consequently, FinOps advocates proactive management strategies, with resource usage prediction emerging as a crucial optimization category. In this research, we introduce the multi-time series forecasting system (MSFS), a novel approach for data-driven resource optimization alongside the hybrid ensemble anomaly detection algorithm (HEADA). Our method prioritizes the concept-centric approach, focusing on factors such as prediction uncertainty, interpretability and domain-specific measures. Furthermore, we introduce the similarity-based time-series grouping (STG) method as a core component of MSFS for optimizing multi-time series forecasting, ensuring its scalability with the rapid growth of the cloud environment. The experiments performed demonstrate that our group-specific forecasting model (GSFM) approach enabled MSFS to achieve a significant cost reduction of up to 44%.

The development of autonomous vehicles (AVs) has seen rapid advancements over the past decade, driven by significant improvements in artificial intelligence, sensor technologies, and computational power. This survey paper provides a comprehensive review of end-to-end perception and prediction systems for autonomous vehicles, which are critical components for enabling safe and efficient navigation in complex environments. We systematically examine the state-of-the-art approaches, focusing on deep learningbased methods that integrate perception (e.g., object detection, semantic segmentation) with prediction (e.g., trajectory forecasting) in a unified framework. The paper explores the key challenges associated with end-to-end learning, such as data fusion from multiple sensors, the tradeoffs between accuracy and computational efficiency, and the interpretability of learned models. Additionally, we review publicly available datasets and benchmarks that have catalyzed progress in this domain. Finally, we discuss the current limitations and potential future research directions, highlighting the need for more robust, scalable, and generalizable solutions to achieve full autonomy in diverse driving scenarios. This survey aims to serve as a valuable resource for researchers and practitioners working towards the next generation of autonomous vehicle technologies.

This study explores the profound impact of emerging technologies on the Zero Trust paradigm and the challenges they present in the evolving cybersecurity landscape. As organizations grapple with increasingly complex threats, the integration of innovative technologies with Zero Trust principles offers both promising solutions and new hurdles. The research investigates how artificial intelligence, machine learning, blockchain, quantum computing, and cloud/edge technologies are reshaping the implementation and efficacy of Zero Trust architectures. These technologies enable more sophisticated trust evaluation algorithms, enhanced threat intelligence, and dynamic access control mechanisms, pushing the boundaries of traditional Zero Trust models. The rapid pace of innovation introduces complexities in maintaining continuous verification and least-privilege access across hybrid and multi-cloud environments. Moreover, the integration of AI and machine learning in Zero Trust frameworks raises questions about data privacy, algorithmic bias, and the need for explainable security decisions. The paper proposes a methodology for addressing these challenges, emphasizing the need for adaptive Zero Trust strategies that can evolve with technological advancements. By examining real-world case studies and conducting empirical research, this study provides insights into the practical implications of emerging technologies on Zero Trust implementation. It offers guidance on how enterprises can harness these technologies to create more resilient, responsive, and effective cybersecurity measures. Ultimately, this review aims to equip organizations with the knowledge and strategies needed to embrace emerging technologies within a Zero Trust framework, enabling them to navigate the complex interplay between innovation and security in the digital age.

Since the launch of 5G services in Thailand in 2020, there have been continuous improvements in 5G coverage. Currently, 5G coverage extends to most areas throughout the country. However, coverage issues persist not only in rural areas but also in high-rise buildings in urban areas. Consequently, a study was conducted within such buildings. This paper assesses the performance of 5G at different altitude test points. The chosen location for the field tests was a high-rise building within a crowded public hospital, which receives numerous patients every weekday, in the major urban area of Bangkok. Two smartphones from the same manufacturer, both supporting 5G technology and equipped with the Speedtest application, were employed as tools for this study. Tests were carried out on the third and twenty-fourth floors of the high-rise building for data collection. The primary finding of this study reveals that download speeds exhibited a significant decrease with increasing altitude of the test points, as evidenced by statistical analysis (pvalues < 0.001). This implies an issue with altitude-induced variations, indicating a need for the improvement of indoor 5G coverage in high-rise buildings.

In this paper, a new control algorithm was developed for a body weight support system (BWS). BWSs are used for rehabilitation purposes and to help patients regain the ability of walking after conditions like stroke or other neurological disorders. For designing the controller, different approaches, model predictive control; classical PID control and LQR control, have been used and the results are compared.

This paper deals with classifying dog behavior using motion sensors, leveraging a transformer-based Deep Neural Network (DNN) model. Understanding dog behavior is essential for fostering positive relationships between dogs and humans and ensuring their well-being. Traditional methods often fall short in capturing temporal dependencies and efficiently processing highdimensional sensor data. Our proposed architecture, inspired by its success in Natural Language Processing (NLP), utilizes the self-attention mechanism of the transformer to effectively identify relevant features across various time scales, making it ideal for real-time applications. The architecture includes only the encoder part with a classifier's head to output probabilities of dog behavior. We used an open-access dataset focusing on seven different dog behavior, captured by motion sensors on top of the dog's back. Through experimentation and optimization, our model demonstrates superior performance with an impressive accuracy rate of 98.5%, outperforming time-series DNN models. The model's efficiency is further highlighted by its reduced computational complexity, lower latency, and smaller size, making it well-suited for deployment in resource-constrained environments.

In the era of Industry 4.0, the proliferation of dataintensive industrial applications poses a tremendous challenge on current edge computing networks. The collaboration of computing powers across standalone edge servers is vital to provision such delay-sensitive services for smart devices (SDs), enabling the converged computing and networking at the edge. In this paper, we propose a new computing paradigm named edge-driven computing power networks (CPNs) by orchestrating the computing and network resources of edge servers through the centralized resource scheduling and decentralized task computing. However, efficient task offloading and collaborative processing is challenging, which requires higher degrees of network automation and intelligence. Therefore, we present an architecture that integrates digital twin (DT) technology into the edge-driven CPNs, where the DTs are created as the digital replicas to assist both the computation offloading and collaborative processing. The objective is to investigate the system average weighted utility maximization through the joint design of the computing power assignment at the edge server as well as the service association, task partition, and transmit power control at the SD. Due to the temporal-spatial variability of tasks and the resulted dynamic environment, we transform the resultant problem as a Markov decision process aiming at maximizing the long-term average weighted utility. To efficiently handle the high-dimensional discrete-continuous action space, we develop a hybrid soft actor-critic based deep reinforcement learning algorithm, which is a continuous-discrete converged variant of soft actor-critic. Simulation results validate the superiority of our proposed algorithm over the benchmarks, and further show the significant gains can be obtained by integrating DTs into the edge-driven CPNs.

1. IntroductionHumanity has taken great leaps to achieve the position that it is in today. Artificial Intelligence (AI) involves the development of machines or equivalent algorithms for performing tasks to reduce redundancy and increase reliability. As is with any other discovery or innovation, AI has evolved to suit the needs of the world it lives in, and its great flexibility gives it the power to be applied across fields, with a constant look for patterns or ideologies that could be used to address any problems.Intelligence has evolved beyond human intelligence, a concept widely and deeply in psychology by pioneers such as Binet, Spearman, Gardner, and many others. Machines and software are now frequently considered more “intelligent” - smarter, faster, more precise. Thus, it is not surprising that psychology and artificial intelligence have contributed to each other. Psychologists have been increasingly involved in the development of AI systems. They have been trying to ensure human biases are not implemented in the systems, among other concerns such as trust, Privacy, and how people will relate to the bots [1].The first AI system developed to act as a therapist potentially was ELIZA, principled after Carl Rogers and his client-centered approach [2]. ELIZA was a text-based chatbot that worked with natural language processing. Advancements have led to more sophisticated software, which may also learn and enhance through human communication, unlike their simpler predecessor. As discussed throughout the paper, the insurance of good mental health has been realized as an important part of any individual, and the aspects related to the same have been deliberated upon in detail. AI’s power allows it to be specifically used to address issues or problems related to mental health counseling and therapy.

In this letter, to handle the over-generation issue of photovoltaic (PV) systems based on maximal power point tracking (MPPT) algorithms, and to provide the power reservation for power grids, an optimal point shifting design based on virtual impedance is proposed for PV systems with flexible power point tracking. With the introduction of the virtual impedance, the virtual PV power is provided to auto-shift optimal power points on PV power-voltage curves. With a MPPT algorithm applied to the virtual PV power, the actual PV power is auto-turned without the estimation of maximum available PV power. The proposed design is validated through simulation studies of a PV-battery charging system and experimental studies of a grid-tied PV system, where both the overcharging of the battery and the over PV power generation are effectively avoided by reducing actual PV power according to the increasing of the battery SOC or the grid RMS voltage, respectively.

The increasing sophistication of ransomware attacks poses significant challenges for organizations attempting to secure their digital infrastructures. Simulating such attacks requires innovative solutions that can adapt to rapidly evolving threats and provide comprehensive testing of security measures. A novel approach utilizing autonomous AI systems for ransomware simulations offers unprecedented scalability and adaptability, significantly enhancing the traditional red teaming process. This paper explores how AI-driven attack simulations generate novel, previously unseen ransomware vectors, evaluating their success rates, stealth capabilities, and effectiveness against various defensive countermeasures. Through a modular architecture, the framework described integrates reinforcement learning to continuously refine attack strategies, ensuring that it remains effective in both simple and complex network environments. The results demonstrate that AI-generated simulations not only match but often surpass the capabilities of traditional red teaming methods, especially in terms of novelty and detection evasion. Moreover, the scalability of this approach allows for continuous, large-scale testing without requiring significant human oversight, making it an ideal solution for organizations aiming to stay ahead of emerging ransomware tactics. Future work could focus on refining the AI models further and expanding their defensive applications to create a fully integrated cybersecurity solution.

The rich-club phenomenon describes the tendency of well-connected nodes to form highly cohesive cores within networks, offering insights into their fundamental structure. This paper presents the Rich-Club Integrated (RCI) model, developed for predicting missing edges in large and complex bipartite networks, which is particularly relevant for product recommendation systems. By incorporating the rich-club coefficient, RCI demonstrates enhanced performance compared to models that do not account for this phenomenon. This study highlights the practical utility of the rich-club coefficient in enhancing recommendation systems and encourages further exploration of this approach in network analysis.