Ransomware is a well-known method that cyberattackers frequently use to infect their victims. This research aims to compare the performance of machine learning techniques, which are the support vector machine, Gradient Boosting, and Random Forest, with respect to Ransomware using two dataset repositories to predict crypto-ransomware and locker-type ransomware. With the UCI dataset, SVM produced generalization errors of 0.5 Kaggle.com . Gradient Boosting performed better than other techniques, achieving training, testing, validation, and AUC levels of 0.998, 0.993, 1.000, and 0.59

This paper examines the integration of a medium-interaction honeypot (Cowrie) with an Intrusion Detection System (IDS) to enhance cyber threat detection within controlled environments. The hybrid approach aims to overcome the limitations of standalone systems, such as high false positives in anomaly-based IDS or the inability of signature-based IDS to detect zero-day threats, by leveraging deception-based logging and real-time traffic analysis. The system was implemented on a virtual Ubuntu Server using open-source tools and subjected to simulated attacks, including brute force, DoS, scanning, and lateral movement. Log data was correlated and filtered using custom Python scripts, and then visualized in real-time via Tmux for monitoring. Results showed improved detection accuracy, reduced false positives, and enhanced context awareness. The study confirms the effectiveness of integrating honeypots and IDS as a proactive security architecture suitable for constrained environments, offering valuable insights for cybersecurity professionals, educators, and organizations seeking adaptive, lightweight, and cost-effective defense mechanisms.

Reconfigurable intelligent surface (RIS)-assisted non-orthogonal multiple access (NOMA) has emerged as a promising technique to enhance spectral efficiency and coverage in fifth- and sixth-generation wireless networks. However, asymmetric indoor propagation conditions characterized by heterogeneous line-of-sight (LoS) and non-line-of-sight (NLoS) links often degrade user fairness. This paper investigates a downlink RIS-assisted NOMA system under the standardized 3GPP indoor office (InH) channel model to address fairness-oriented design under realistic link-budget constraints. We formulate an optimization problem for max-min fairness that jointly considers discrete RIS element partitioning and NOMA power allocation to achieve a symmetrical allocation of quality of service (QoS). To enable efficient computation, the non-convex problem is transformed into an epigraph form and solved using a low-complexity, bisection-based quasi-convex optimization framework combined with enumeration over RIS partitions. Numerical results demonstrate significant fairness gains; for instance, doubling the RIS array size yields a substantial improvement in the ergodic max-min rate, corresponding to approximately a 66% gain at moderate transmit power levels. Furthermore, by accounting for practical impairments such as imperfect successive interference cancellation (iSIC), imperfect channel state information (iCSI), and RIS implementation losses, the results reveal that fairness-optimal operation consistently prioritizes the far user to overcome severe indoor NLoS attenuation. The proposed framework is also compared with alternating optimization (AO)-based RIS-NOMA, conventional RIS beamforming without partition and RIS-assisted orthogonal multiple access (OMA) schemes. Simulation results confirm that the proposed framework achieves low computational complexity, making it suitable for practical indoor wireless environments.

Smart cities increasingly rely on Artificial Intelligence (AI) and Machine Learning (ML) to enhance urban infrastructure, particularly in transportation and energy systems. This paper investigates AI/ML-driven solutions for optimizing traffic flow, smart parking, and autonomous mobility, as well as improving smart grid efficiency, demand-response management, and renewable energy integration. The study highlights how these technologies enable real-time decision-making, predictive analytics, and operational resilience. Despite these advancements, challenges remain, including data privacy, cybersecurity, scalability, and algorithmic transparency. This paper synthesizes current approaches, identifies critical research gaps, and outlines future directions, including adaptive AI/ML models, integration with emerging technologies such as blockchain, and the development of holistic, resilient smart city architectures. The findings aim to guide both researchers and practitioners in advancing sustainable and intelligent urban infrastructure.

Network slicing is a pivotal element in 5G/6G networks, enabling service isolation and customization over shared infrastructure. This paper addresses the optimization of throughput, latency, and resource usage across three service types: enhanced Mobile Broadband (e-MBB), Ultra-Reliable Low-Latency Communication (URLLC), and massive Internet of Things (m-IoT). The varying requirements of these services pose challenges: e-MBB demands high data rates, URLLC requires ultra-low latency, and m-IoT requires massive connectivity and substantial bandwidth. A Reinforcement Learning (RL) framework is introduced to optimize central processing unit (CPU), memory, and bandwidth (BW) allocations under diverse traffic conditions. The framework jointly optimizes slice admission, dynamic resource allocation, and Quality of Service (QoS) guarantees in a cloud-native network. A synthetic dataset simulating slice demand patterns and traffic dynamics is used to evaluate machine learning models, including supervised learning and RL. Results show that m-IoT consumes the most bandwidth, e-MBB experiences the highest latency, and URLLC maintains stable, low latency. RL-based allocation improves efficiency, reducing service violations and supporting 5G/6G network scalability. Findings validate the efficiency of the proposed approach, demonstrating significant improvements in latency, throughput, and resource efficiency compared to fixed-rule baselines. This system emphasizes adaptive resource management for sustainable network growth.

Breast cancer is a critical health issue where timely diagnosis is essential for treatment and healthier outcomes. Advancements in computational methods now enable real-time breast cancer diagnosis, providing accurate and timely results that significantly reduce patient anxiety and wait times. This study reviews the literature on real-time breast cancer diagnosis and proposes a deep convolutional neural network algorithm with quantization and pruning to enhance diagnostic efficiency and speed. Additionally, the study introduces the Accuracy Disparity Ratio (ADR) as a novel metric to evaluate model bias and measure fairness in diagnostic outcomes across different patient groups. In assessing the model, the proposed ADR compares model performance across different data groups. This normalizes model accuracy by the number of cases in each data group, highlighting differences in performance between groups to ensure that the developed real-time breast cancer diagnostic system will be effective and can be adopted in clinical diagnosis. The proposed decision support system, deployed on edge computing platforms, handles the local processing of real-time breast image data to reduce latency. The proposed system aims to enhance the potential of AI and advanced imaging techniques for real-time breast cancer diagnosis, ultimately improving patient outcomes and diagnosis.

Educational institutions in developing regions face challenges in monitoring student engagement due to resource scarcity, resulting in suboptimal learning outcomes and limited interaction between teachers and students. This paper proposes an affordable, low-power smart classroom monitoring system that leverages TinyML technology to deliver real-time insights into student attentiveness on low-power edge devices. Using TinyML frameworks such as TensorFlow Lite and Edge Impulse, the system performs facial recognition and object detection tasks on energy-efficient microcontrollers, including the Raspberry Pi and Arduino. Unlike traditional cloud-based solutions, this approach ensures low latency, enhanced privacy, and offline functionality, making it ideal for regions with limited internet connectivity. The system categorizes students’ engagement levels into attentive, partially attentive, or inattentive, enabling educators to provide personalized learning interventions. The alignment of the solution with the Sustainable Development Goals, particularly SDG 4, such as Quality Education, and SDG 9, like Industry, Innovation, and Infrastructure, enables scaling up in several educational settings, thereby enhancing learning outcomes for these underserved communities. This work contributes to the advancement of sustainable, affordable, and scalable AI-driven education solutions, thereby fostering inclusive learning experiences worldwide. A pilot study in rural classrooms will validate the system's effectiveness, while collaboration with educational stakeholders will ensure the system's adaptation for broader adoption. The expected outputs of this project are improved academic performance, increased student engagement, and a scalable framework for intelligent learning environments.

The proliferation of connected objects in the Internet of Things (IoT) ecosystem presents challenges in enabling real-time intelligence while safeguarding data privacy, especially within the computational and energy limitations of edge devices. This study, based on simulation and synthetic data collection methods, addresses these challenges by proposing a lightweight edge AI framework that employs federated learning, enabling model training across distributed Internet of People and Things (IoP) devices without transferring raw data to centralised servers. The framework integrates on-device inference, stochastic local updates, and model compression to ensure low-latency decision-making while adhering to memory and energy constraints. To enhance security, differential privacy mechanisms, encrypted aggregation, and robust outlier detection are utilized to defend against adversarial and Byzantine attacks. The proposed framework offers an effective solution for deploying federated intelligence on resource-constrained IoP devices, enabling responsive, privacy-preserving, and resilient edge AI operations in distributed environments.

ABSTRACT In this paper, the secrecy energy efficiency (SEE) maximization problem is investigated in a dual‐reconfigurable intelligent surface (RIS)‐assisted downlink multiple‐input single‐output (MISO) system with an intermittently active mobile eavesdropper under statistical channel state information (CSI). The proposed framework models the eavesdropper channel using a statistical mean‐covariance representation, rather than conventional approaches that assume instantaneous eavesdropper CSI. A deterministic secrecy‐rate surrogate is obtained based on the expected eavesdropper signal‐to‐interference‐plus‐noise ratio (SINR) and an uncertainty‐aware robustness penalty. The resulting SEE optimization problem is nonconvex due to the fractional objective function, coupled beamforming and RIS phase variables, and unit‐modulus RIS constraints. To address this issue, a low‐complexity alternating optimization algorithm is designed using Dinkelbach fractional programming. The results demonstrate that the proposed dual‐RIS statistical‐CSI framework achieves 9.62 Mbit/J SEE at 30 dBm transmit power, outperforming the single‐RIS statistical‐CSI benchmark while approaching the perfect‐CSI dual‐RIS benchmark. Overhead‐aware evaluation further confirms the effectiveness of the proposed dual‐RIS framework under practical operating conditions.