Information Technology University

The rapid advancement of digital technologies, particularly the Internet of Things (IoT) and quantum computing (QC), is reshaping modern military operations by enhancing real-time data collection, autonomous decision-making, and situational awareness. IoT-enabled systems, such as UAVs, battlefield sensors, and smart logistics, provide military forces with operational superiority but also introduce significant cybersecurity risks, including espionage, cyber sabotage, and unauthorized access. The emergence of QC further complicates this landscape, as quantum algorithms can break traditional encryption protocols, threatening the security of military communications and IoT infrastructure. However, quantum technologies also offer revolutionary cybersecurity solutions, such as Quantum Key Distribution (QKD) and Post-Quantum Cryptography (PQC), which leverage the principles of quantum mechanics to ensure secure communication. Despite these advancements, challenges such as the technological maturity of quantum solutions, high implementation costs, and global competition in quantum research pose obstacles to widespread military adoption. This chapter explores the interplay between QC and military IoT security, addressing both the vulnerabilities of IoT systems and the potential of quantum-enhanced security frameworks. By integrating quantum-safe cryptographic techniques, defence organizations can fortify military IoT infrastructure, mitigate cyber threats, and maintain strategic superiority in an era of increasing digital warfare.

• ConVLM addresses the limitation of coarse alignment in existing VLMs by introducing context-guided token learning and enhancement, enabling fine-level image-text interaction that captures subtle morphological details in histology images. • The model selectively removes irrelevant visual tokens and enhances relevant ones through integrated modules across encoder layers, resulting in richer and more discriminative visual representations for downstream tasks. • ConVLM is trained using a novel context-guided token learning loss, which guides the model to focus on contextually important tissue structures, improving generalization and interoperability. • Evaluated on both ROI-level and WSI-level classification tasks, ConVLM outperforms SOTA models, demonstrating robust generalization across diverse histopathology datasets and tasks such as cancer subtype prediction and survival analysis. Vision-Language Models (VLMs) have recently demonstrated exceptional results across various Computational Pathology (CPath) tasks, such as Whole Slide Image (WSI) classification and survival prediction. These models utilize large-scale datasets to align images and text by incorporating language priors during pre-training. However, the separate training of text and vision encoders in current VLMs leads to only coarse-level alignment, failing to capture the fine-level dependencies between image-text pairs. This limitation restricts their generalization in many downstream CPath tasks. In this paper, we propose a novel approach that enhances the capture of finer-level context through language priors, which better represent the fine-grained tissue morphological structures in histology images. We propose a Context-guided Vision-Language Model (ConVLM) that generates contextually relevant visual embeddings from histology images. ConVLM achieves this by employing context-guided token learning and token enhancement modules to identify and eliminate contextually irrelevant visual tokens, refining the visual representation. These two modules are integrated into various layers of the ConVLM encoders to progressively learn context-guided visual embeddings, enhancing visual-language interactions. The model is trained end-to-end using a context-guided token learning-based loss function. We conducted extensive experiments on 20 histopathology datasets, evaluating both Region of Interest (ROI)-level and cancer subtype WSI-level classification tasks. The results indicate that ConVLM significantly outperforms existing State-of-the-Art (SOTA) vision-language and foundational models. Our source code and pre-trained model is publicly available on: https://github.com/BasitAlawode/ConVLM

Purpose This study aims to systematically review the methods, predictive factors and evaluation metrics used to forecast university students’ academic performance. By synthesizing recent literature, it addresses the lack of an integrated overview of state-of-the-art approaches and key determinants, and provides evidence-based insights to inform personalized teaching strategies and early intervention practices in higher education. Design/methodology/approach A systematic literature review (SLR) was conducted in accordance with the PRISMA framework. Articles published between 2015 and 2024 were retrieved from the Web of Science database using predefined keywords and Boolean search strategies. Following a structured screening process based on explicit inclusion and exclusion criteria, 31 studies were selected for analysis. These studies were systematically examined to synthesise predictive methods, key influencing factors and commonly used evaluation metrics in forecasting university students’ academic performance. Findings Academic performance is influenced by personal factors (e.g. entry scores, prior Grade Point Average [GPA], gender, age, emotional responses, self-esteem) and external factors (e.g. family income, parental education, accommodation, learning environment). Main prediction approaches include statistical regression models, machine learning algorithms (e.g. random forest, support vector machine (SVM), gradient boosting) and deep learning techniques. Machine learning shows strong performance with complex, non-linear data, while deep learning achieves high accuracy but requires large datasets. Common evaluation metrics include accuracy, precision, recall, F1 score and mean squared error (MSE). Originality/value This study provides a systematic and integrative synthesis of predictive methods, influencing factors and evaluation metrics related to university students’ academic performance. By consolidating evidence that has previously been scattered across methodological and thematic strands, the review offers a structured reference for researchers and actionable insights for educators and policymakers to support early identification of at-risk students and inform data-driven educational decision-making.

Decentralized anonymity networks such as I2P rely on a core assumption: that key algorithms (such as peer selection) behave the same across all participating nodes. We show this assumption breaks down when the network runs multiple different software

Polyp detectors trained on clean datasets often underperform in realworld endoscopy, where illumination changes, motion blur, and occlusions degrade image quality. Existing approaches struggle with the domain gap between controlled laboratory conditions and clinical practice, where adverse imaging conditions are prevalent. In this work, we propose ADAPTIVEDETECTOR, a novel two-stage detector-verifier framework comprising a YOLOv11 detector with a vision-language model (VLM) verifier. The detector adaptively adjusts per-frame confidence thresholds under VLM guidance, while the verifier is fine-tuned with Group Relative Policy Optimization (GRPO) using an asymmetric, cost-sensitive reward function specifically designed to discourage missed detections-a critical clinical requirement. To enable realistic assessment under challenging conditions, we construct a comprehensive synthetic testbed by systematically degrading clean datasets with adverse conditions commonly encountered in clinical practice, providing a rigorous benchmark for zero-shot evaluation. Extensive zero-shot evaluation on synthetically degraded CVC-ClinicDB [1] and Kvasir-SEG [2] images demonstrates that our approach improves recall by 14 to 22 percentage points over YOLO alone, while precision remains within 0.7 points below to 1.7 points above the baseline. This combination of adaptive thresholding and cost-sensitive reinforcement learning achieves clinically aligned, open-world polyp detection with substantially fewer false negatives, thereby reducing the risk of missed precancerous polyps and improving patient outcomes.

The rapid proliferation of IoT devices in smart environments necessitates scalable, user-friendly management systems. However, current platforms often lack integration between conversational agents and realtime Digital Twin (DT) supervision. This paper proposes a novel architecture that integrates the Model Context Protocol (MCP) with Home Assistant (HA) to manage a 3D DT dashboard for a smart campus for better energy consumption using a natural language interface powered by Cursor. The system employs an intermediary component, “mcp-proxy”, which bridges the MCP client (Cursor) and the HA MCP server, allowing bidirectional communication over standard input/output and SSE protocols. Through natural language queries, users can monitor and control IoT devices represented within a 3D DT model. The deployment includes smart equipment in the lab, kitchen and mailroom, demonstrating the complete data flow from user interaction to structured response. A case study on energy consumption monitoring illustrates how the proposed system can be used in practical applications. Experimental evaluations with various LLMs confirm the feasibility, responsiveness, and contextual accuracy of the proposed solution. This work highlights the potential of agentic LLMs and open protocols in enabling intuitive and intelligent management of smart sustainable environments.

Conventional wireless communication systems are primarily dependent on the radio frequency spectrum, which will gradually become congested and bandwidth-limited due to increasing global internet traffic. Optical wireless communication, coupled with structured light, is the key to providing tariff-free, high-speed internet. Orbital angular momentum in structured light beams can facilitate multiplexing of spatial modes. Optical vortex beams have been widely studied for communication, but they suffer from beam divergence, sensitivity to turbulence, and constraints on receiver aperture. Hermite-Gaussian (HG) beams are a compelling alternative, featuring orthogonality, lower crosstalk, and compact beam profiles in low-order modes, making them ideal for efficient optical wireless communication. However, conventional optics for HG beam generation introduce bulkiness and system complexity, limiting their integration in photonic systems. A single-cell-driven, broadband metasurface platform is numerically studied to generate HG beams in the visible wavelength range (488−633 nm), addressing the aforementioned issues. The proposed metasurfaces, made of rectangular-shaped zinc sulfide unit cells, achieve an average transmission efficiency of 80% across the visible spectrum. These advancements could enhance optical trapping, tweezers, and non-invasive handling, while also innovating free-space optical communication, multiplexed communication, and structured illumination in biomedical imaging.

As the most aggressive skin malignancy and a rapidly spreading cancer worldwide, the early detection of melanoma is crucial. This paper introduces an advanced hybrid model for classifying melanoma in dermoscopic images, building on the success of deep convolutional neural networks (CNNs) over traditional algorithms. We propose a system that evaluates deep features extracted from four distinct models (ShuffleNet, Darknet53, Xception, and SqueezeNet), which are then fed to two proven machine learning classifiers, Random Forest (RF) and Support Vector Machine (SVM). The models were trained and tested on 3,297 images from the International Skin Imaging Collaboration (ISIC) public dataset. The best-performing configuration, utilizing DarkNet53 features with an RF classifier, achieved 88.01% accuracy, 87.69% sensitivity, and 88.09% specificity, surpassing the performance of existing state-of-the-art solutions on the same data.