Proceedings of the 2nd Symposium on Smart, Sustainable, and Secure Internet of Things

Inhaltsverzeichnis

1. Frontmatter

   PDF-Version jetzt herunterladen

2. The Next Frontier of Cybersecurity: Zero Trust for Enterprise IoT Ecosystems

   • Open Access

   Belal Ali, Omar Amjad Dib

   Abstract

   The rapid proliferation of Internet of Things (IoT) devices across enterprise environments has significantly expanded the digital attack surface, exposing organisations to new classes of security threats. Traditional perimeter-based security models are increasingly ineffective in this context, as IoT devices often lack standardised controls, operate in distributed ecosystems, and frequently interact with untrusted networks and services. This paper presents a structured Zero Trust Security (ZTS) framework explicitly tailored for enterprise IoT ecosystems. Grounded in the principles of continuous verification and least-privileged access, the proposed five-step methodology addresses the unique security challenges posed by IoT deployments. Key components include protect surface definition, transaction flow mapping, Zero Trust network architecture, granular policy enforcement, and continuous monitoring. The paper also identifies practical implementation challenges, such as device discovery, authentication, and segmentation, and offers guidance on over- coming them. By aligning ZTS principles with IoT-specific requirements, this work contributes a scalable and adaptable approach to building resilient, context-aware security architectures for future enterprise infrastructures.

   PDF-Version jetzt herunterladen

3. Next-Generation IoT Connectivity: FSO-RF Integration with High-Altitude Platforms for Reliable Data Transmission

   • Open Access

   Ramy Samy, Shuang Li, Fang Xu, Basem M. ElHalawany

   Abstract

   The increasing demand for high-speed and reliable Internet of Things (IoT) connectivity presents serious challenges to existing terrestrial and satellite-based networks, such as congestion, high costs, and scalability limitations. High-altitude platforms (HAPs) are a viable alternative, providing wide-area coverage and low-cost deployment. However, they are constrained by limited radio frequency (RF) bandwidth. We propose a hybrid free-space optical (FSO)-RF communication system for HAP-enabled IoT connectivity to address this limitation. FSO can cover a dense area of IoT communications at high speed through FSO links to access points (APs). At the same time, RF can provide wide area coverage to IoT devices outside the range of APs. We analyze the system’s performance by deriving the ergodic capacity for both APs and remote IoT devices. Through Monte Carlo simulations, we validate the theoretical model, while numerical results demonstrate the system’s effectiveness in enhancing IoT connectivity.

   PDF-Version jetzt herunterladen

4. Multimodal Pain Recognition: Integrating Facial Expressions and Biomedical Signals with Deep Learning

   • Open Access

   Arhina Ghosh, Neha Tyagi, Nitin Rakesh, Balamurugan Balusamy, Akhil Gupta

   Abstract

   Pain expression is characterized by high complexity and multidimensionality. It often uses subjective self-reporting as precise pain assessment remains an essential problem. In this research work, a new methodology for recognition of pain is represented based on a hybrid model that combines convolutional neural networks (CNNs) and particle swarm optimization (PSO) techique along with bidirectional long short- term memory (BiLSTM) networks. This study used video and physiological signals as input data. Experimental results significantly improved the binary pain classification accuracy over several existing models. This study is giving results of 86.3% accuracy and proving the hybrid approach to be very effective. The CNN handles extraction of spatial hierarchies from video data, while the PSO-optimized BiLSTM appropriately models the temporal evolution of physiological responses. Through multimodal integration and advanced optimization approaches, a pathway is paved toward a more reliable and efficient pain recognition model which will eventually help in improving patient care.

   PDF-Version jetzt herunterladen

5. Detection of Diabetic Macular Edema Using Deep Learning: A Comparative Study

   • Open Access

   Ismaeel Al Ridhawi, Ali Abbas, Hasan Fayyad-Kazan

   Abstract

   Diabetic Macular Edema (DME) is one of the major causes of blindness in adults, primarily due to fluid accumulation in the macula derived from Diabetic Retinopathy (DR). In this paper, a Deep Learning (DL)-based diagnostic framework is developed to detect and identify various levels of DR severity, thereby facilitating early identification and management of DME. Our methodology begins with a comparison of several convolutional neural network (CNN) architectures for baseline modeling, including VGG 13, VGG 16, Resnet 18, Resnet 101, ResNeXt 50, EfficientNet B0, EfficientNet B3, EfficientNet B4, and ShuffleNet. Among these, the VGG-13 model was found to be the best performing model with an accuracy of 84.6% in identifying the severity of DR. Based on these outcomes, we introduce a Mixture of Experts (MoE) model that combines the capabilities of several task-specific models to increase the overall classification quality. This ensemble-based method improves the diagnostic accuracy to 93.7%, representing a 9.1% improvement over the baseline. These results underscore the potential of MoE in providing reliable DME diagnosis.

   PDF-Version jetzt herunterladen

6. Enabling Low-Cost Access to Weather Satellite Imaging: A Ground Station Solution for IoT Applications

   • Open Access

   Ramy Samy, Ahmed Sakr, Mohamed Hassan, Hassan Ahmed, Basem M. ElHalawany

   Abstract

   Polar satellites play an important role in gathering information on various weather phenomena. However, real-time access to satellite imagery remains constrained by operational costs and the technical complexity of receiving ground stations (GS). This paper presents the design and implementation of a low-cost, portable, and easily deployable GS to receive signals from weather satellites operating in the VHF band. The GS integrates standard commercial hardware with open-source software for receiving and processing satellite images. We perform detailed link budget calculations using the System Tool Kit (STK) software to ensure reliable data transmission during communication sessions. Experimental validation is demonstrated through the successful acquisition of NOAA-19 weather images. The results highlight a cost-effective and practical solution that improves access to real-time weather information for a wider community and Internet of Things applications.

   PDF-Version jetzt herunterladen

7. Drone-Aided Agriculture 5.0: A Survey on Machine-Learning and IoT Paradigms for Smart Farming

   • Open Access

   Basem M. ElHalawany, Shahad Alshammeri, Sara Aldaihani, Manar Alyouhah, Rahaf Alrashidi

   Abstract

   Agriculture 5.0 leverages emerging technologies such as drones, Internet-of-Things (IoT), machine-learning (ML), and digital twins to enhance precision farming and improve crop health management. Drone-based monitoring combined with advanced ML algorithms and IoT sensors has become a crucial tool for real-time agricultural surveillance. This survey explores the integration of Unmanned Aerial vehicles (UAVs) in smart farming. We review various drone-mounted sensor technologies and payloads for plant health assessment and environmental monitoring. Furthermore, we analyze state-of-the-art ML techniques, including supervised, unsupervised, reinforcement, and federated learning, for plant classifications, disease detection, yield estimation, anomaly identification, and privacy preservation.

   PDF-Version jetzt herunterladen

8. Adaptive Hyper-Heuristics for Smart Logistics Optimization

   • Open Access

   Kassem Danach, Hasan Fayyad-Kazan, Wissam Khalil, Samir Haddad, Jinane Sayah

   Abstract

   The complexity of logistics combinatorial optimization problems including vehicle routing, warehouse scheduling, and dynamic delivery has increased because of rising demand and evolving constraints. Metaheuristics show effectiveness but need problem-specific tuning and demonstrate limited general applicability. This research presents a learning-based hyper-heuristic framework which operates at high abstraction levels to select or generate low-level heuristics through dynamic decision-making based on problem features and real-time performance feedback. The proposed system uses reinforcement learning to select heuristics while pursuing adaptability, scalability and domain independence. Additionally, the framework demonstrates its effectiveness through benchmark dataset experiments, which show better solution quality and improved computational efficiency and robustness compared to traditional metaheuristics. Moreover, the framework shows its capability to perform automated decision-making while minimizing human involvement and demonstrating effective adaptation to changing logistics environments. Finally, this research presents an adaptable intelligent optimization system which enhances operational efficiency and resilience in smart supply chain systems.

   PDF-Version jetzt herunterladen

9. Metaverse-Driven Immersive Therapy: Reducing Air Travel Anxiety for Individuals with Autism

   • Open Access

   Mariam Thaher, Shurouq Khan, Fatima Alqattan, Tala Alhammouri, Ismaeel Al Ridhawi

   Abstract

   Virtual reality (VR) functions as a substantial transformative tool which delivers therapeutic interventions particularly in exposure therapy. This paper investigates the implementation of metaverse-based virtual reality to create flight simulations for children with autism spectrum disorder (ASD) and social phobia. The proposed solution provides a personalized controlled space through which users can experience airplane boarding, in-flight interactions, and disembarkation with the support of VR tools. The system aims to expose users to these situations within a controlled space to minimize their stress levels before they face actual air travel. The system would adapt the environment to meet the specific needs of the user, thus offering an individualized therapeutic approach. This adaptation is performed with the aid of visual and audio feedback by incorporating verbal and a non-verbal user behavior determination models. The models would identify the user’s attitude through speech analysis, facial recognition, physiological reactivity identification, and gesture recognition. The system is a work-in-progress and is expected to offer a personalized and real-time adaptive approach once complete, making it more effective than other methods in the literature.

   PDF-Version jetzt herunterladen

10. CLAHE-Augmented MRI Dementia Classification via Soft-Voting Ensemble with Gradient-Based Analysis and Dementia Correlation

    • Open Access

    Garima Shukla, Vanshaj Awasthi, Prashant Dubey, Sachin Kumar, Balamurugan Balusamy, Ahmad Alkhayyat

    Abstract

    Dense and accurate automated dementia stage classification from MRI scans is still a challenging task because of the inconsistencies in structural pat- terns and the inability of current deep learning models to manage both local and global feature representations effectively. In this work, a new hybrid ensemble model is proposed that combines CNN-based (ResNet-50, ConvNeXt, DenseNet-121) and transformer-based (Swin Transformer, ViT) models to synergistically extract their complementary strengths for effective dementia classification. An explicitly crafted preprocessing pipeline, including N4 bias field correction, CLAHE, non-local means denoising, Sobel edge detection, and K-means segmentation, provides high-quality input for enhanced feature extraction. The model was tested on a dataset of 87,000 MRI scans with 8500 images independently tested. Experiment results show Swin Transformer obtained AUC-ROC scores of 0.99 for Mild Dementia and 0.97 for Moderate Dementia, and the ensemble model delivered balanced decisions for all stages of dementia. Cohen’s Kappa (0.57) and Matthews Correlation Coefficient (0.60) established model reliability, and class-wise precision-recall substantiated its discrimination capacity. Furthermore, Grad-CAM and attention maps enabled model interpretability by emphasizing primary regions driving predictions, thereby improving clinical relevance. The results highlight the effectiveness of transformer-CNN ensembles in neuroimaging-based dementia diagnosis and promise their suitability for use in real-world diagnostic assistance.

    PDF-Version jetzt herunterladen

11. An IoT-Based Framework for Adaptive Fitness Recommendations Using Clustering and Ensemble Regression

    • Open Access

    Naween Kumar, Ayushman Pranav, Ankit Dubey, Rajesh Kumar Modi, Firoz Khan, Ahmad Alkhayyat

    Abstract

    Personalizing fitness recommendation systems is important to engage users of wearables and maximize their health outcomes. This research provides a new framework for the Internet of Things (IoT), which uses unsupervised clustering, collaborative filtering, and ensemble regression to make adaptive fitness recommendations. The methodology begins by grouping tracker activities and generating characteristics such as activity intensity and calorie expenditure per step. K-Means clustering helps to categorize users into five categories according to their activity level, from inactive to very active, for tailor-made recommendations. The lesson plans are adjusted according to the similarity in behavior using a collaborative filtering algorithm based on Nearest Neighbors (KNN) similarity. The random forest regression achieves near-perfect accuracy in predicting calorie burn (R² = 0.9997), outperforming other regression models. In addition, the system simulates a real-time adjustable recommendation, recommending that users change the intensity of their exercise based on a predicted versus actual calorie deficit. Visual aids and the accuracy of the regression confirm the separation of clusters with different activity patterns. In contrast, qualitative analysis enhances the model’s ability to identify and distinguish different activity patterns. The results demonstrate the framework’s effectiveness in synthesizing user segmentation, predictive analytics, and dynamic recommendation and provide a solid basis for future innovations in real-time train adaptation.

    PDF-Version jetzt herunterladen

12. IoT-Driven Urbanization and Economic Forecasting with Machine Learning

    • Open Access

    Naween Kumar, Ayushman Pranav, Ankit Dubey, Firoz Khan, Ahmad Alkhayyat

    Abstract

    Economic development and urbanization depend highly on traditional metrics that provide little geographic specificity and delayed reporting times that hinder continuous monitoring. This paper proposes a method to automate the prediction of economic conditions by machine learning models, together with the Internet of Things (IoT) sensors and satellite observations of night lights. Combining a refined data pipeline and its stages, including data transformation, with the generation of features and clustering, followed by predictive modeling, provides robust analysis. New indicators such as light density and light per person support the spatial and temporal analysis. The K-means algorithm distinguishes five clusters ranging from entirely rural to completely urban. Regression models, including linear regression with random forest and gradient enhancement, and XGBoost and CatBoost yield a r² of 95.21%. The suggested work provides evidence of the increased urbanization and changes in social structures and political developments. The method of analysis speeds up and improves the accuracy of forecasting urbanization patterns in conjunction with economic developments, thus providing valuable information for urban planners and policymakers.

    PDF-Version jetzt herunterladen

13. Hybrid Neural Network for Stock Prediction: Integrating Multi-source Data with Particle Swarm Optimization

    • Open Access

    Harmanjeet Singh, Rupali Dhir, Chander Prabha, Balamurugan Balusamy, Mahmoud Ahmad Al-Khasawneh, Firoz Khan

    Abstract

    This study presents a novel approach to predicting stock prices by integrating a hybrid deep learning model with a comprehensive analysis of technical data. The methodology utilises Python and its associated modules to systematically acquire, process systematically, and model financial time series data. Historical stock data for blue-chip stocks from the Bombay Stock Exchange (BSE) is obtained, and various technical indicators, such as volume, volatility, trend, and momentum, are calculated. A hybrid deep learning architecture employing long-short-term memory (LSTMS) and a multi-head attention mechanism captures complex temporal correlations in the technical indicator data. The model’s predictive capability is comprehensively evaluated, and the findings illustrate the effectiveness of the proposed strategy in forecasting stock prices. The framework achieved a maximum R² of 0.97 and a minimum MAPE of 1.67 for stock LT when the test data was predicted over 200 epochs with a learning rate of 0.001. Nonetheless, the framework achieved an accuracy of 89% when trained on 5000 trading days, exceeding the performance of those trained on fewer than 5000 days.

    PDF-Version jetzt herunterladen

14. A Movie Recommendation Model Integrating Viewer’s Emotional Experience and TV Feature Extraction

    • Open Access

    Harmanjeet Singh, Chander Prabha, Preeti Sharma, Balamurugan Balusamy, Ahmad Alkhayyat, Nithya Rekha Sivakumar

    Abstract

    Acceptance of information-gathering behaviours, especially from the viewpoints of others, is crucial for enhancing decision-making processes. In film critiques, audience feedback offers significant insights regarding a movie’s quality and value as a time investment. The growing volume of review data requires automation for effective processing. This study introduces the creation of a sophisticated movie recommendation engine that amalgamates many data sources, such as IMDB movie reviews, Netflix ratings, YouTube trailer interactions, and Twitter discourse. User-generated input, including comments, likes, tweets, and trailer replies, is integrated to improve the recommendation process. The suggested methodology utilizes collaborative and content-based filtering, integrating users’ social influence as assessed through their Twitter activity and social characteristics. A hybrid recommendation method produces an initial array of film recommendations. Thereafter, sentiment analysis is utilized to enhance and improve the recommendations. The system employs artificial intelligence methodologies, utilizing the IMDB dataset to train and validate a BERT embedding layer alongside a Bi-Directional Long Short-Term Memory (LSTM), a Bi-Directional Gated Recurrent Unit (GRU) incorporating a self-attention mechanism, and a Convolutional Neural Network (CNN). The model attained a testing accuracy of 93.91% and an AUC of 0.9831, indicating exceptional performance in binary sentiment categorization relative to current methodologies.

    PDF-Version jetzt herunterladen

15. IoT for Next-Generation Networks: A Succinct Review

    • Open Access

    Shalini Kumari, Chander Prabha, Balamurugan Balusamy, M. A. Al-Khasawneh, Firoz Khan

    Abstract

    The analysis predicts that over half a trillion Internet of Things (IoT) devices will become connected during the next decade as automotive sectors strive to enhance 90% of their vehicles through IoT capabilities. The connected vehicle market surpassed 250 million units in 2020, with a 67% growth rate. Integrating IoT technology into industrial operations creates new business prospects that let organizations advance their methods and achieve optimized operations. Technological advancements lead to a smarter world where users gain easier resource access and have improved quality of service (QoS) and quality of experience (QoE). The paper presents six essential factors for IoT, including identification protocols and calculation functionalities that create operational boundaries. The IoT leads industrial operations into transformation, while research between engineering and social sciences needs to develop the technology of connected systems.

    PDF-Version jetzt herunterladen

16. Training Efficiency of DDQN-Based Multilevel Inverter Control: The Influence of Reward Function Penalty Terms

    • Open Access

    Alamera Nouran Alquennah, Sara Hamed, Tassneem Zamzam, Haitham Abu-Rub, Mohamed Trabelsi, Sertac Bayhan, Ali Ghrayeb, Sunil Khatri

    Abstract

    Reinforcement Learning (RL)-based controllers have recently gained attention as AI-driven, model-free methods for controlling power electronic converters by learning optimal control actions through continuous interaction with the environment. Their learning process is governed by a reward function, which guides the agent’s behavior. This paper investigates the influence of incorporating penalty terms into the reward function on the training efficiency and performance of an RL-based controller for a 7-level grid-tied Packed-U-Cell (PUC7) multilevel inverter. The controller is developed using the Double Deep Q-Network (DDQN) algorithm, selected for its balanced combination of strong performance and ease of implementation. The control objectives include sinusoidal current injection into the grid and capacitor voltage regulation around the desired value. The reward function is designed based on current and voltage tracking errors, with two penalty terms introduced to limit deviations beyond predefined thresholds. The study evaluates the impact of varying these penalty magnitudes on learning speed, convergence behavior, and tracking quality. Simulations are conducted in MATLAB/Simulink, demonstrating that the appropriate selection and application of penalties improve training efficiency without compromising control performance.

    PDF-Version jetzt herunterladen

17. IntelliGuard: IoT-Enabled Autonomous Spybot Intelligence for Real-Time Surveillance in Next-Generation Security Applications

    • Open Access

    Dhruv Dhayal, Pratham Aggarwal, Manzoor Ansari

    Abstract

    The rapid advancements in ubiquitous computing have made surveillance robots vital in military and security systems. IoT and sensor miniaturization enable new autonomous monitoring capabilities where traditional fixed installations or human operators prove ineffective. IntelliGuard, an ESP-32-based IoT surveillance robot, integrates camera streaming, motor-driven navigation, and centralized processing for military applications. The system provides autonomous mobility, real-time surveillance, and secure data transmission while remaining cost-effective. Radar and ultrasonic sensors enable precise obstacle detection within three meters. Field testing revealed performance variations under different environmental conditions, with optimal functionality in low-interference settings. Response latency remains minimal while IoT connectivity enables remote operation. Performance directly correlates with sensor precision and connection quality. Future developments will focus on more accurate sensors, stronger communication protocols, improved navigation for irregular obstacles, and enhanced filtering algorithms to minimize interference effects. This study establishes a foundation for advanced IoT-connected robotic surveillance systems in military applications.

    PDF-Version jetzt herunterladen

18. An Optimized Learning Framework for Stock Market Prediction Using Multi-source Data and Grey Wolf Optimizer

    • Open Access

    Harmanjeet Singh, Rupali Dhir, Chander Prabha, Balamurugan Balusamy, Ahmad Alkhayyat, Nithya Rekha Sivakumar

    Abstract

    The prediction of stock prices is an evolving domain, especially in emerging nations such as India, where it substantially influences governmental, commercial, and individual investments. The precision of traditional statistical and econometric models diminishes due to the difficulties encountered in handling non-stationary financial time series data. This matter is resolved using machine learning techniques that utilize historical stock prices to enhance predictions. Debt or equity financing, encompassing dual registration on domestic and international exchanges, is how corporations procure capital in global markets. Precise stock price forecasts are essential for educated decision-making and financial progress in global markets, as investors in these equities enhance capital flow. The proposed system utilizes an iterative optimization procedure to seek optimal solutions in the hybrid LSTM model by dynamically adjusting the placements of wolves throughout the search space. The performance of the proposed model is quantitatively evaluated using statistical measures, including the accuracy, coefficient of determination R², Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE).

    PDF-Version jetzt herunterladen