Computer Science علم الحاسوب

Permanent URI for this collection

https://dspace.alquds.edu/handle/20.500.12213/1091

Browse

Now showing 1 - 5 of 80

Developing an Energy-Efficient Hybrid Intrusion Detection System for Wireless Sensor Networks
(Al-Quds Univeersity, 2025-12-20) Murad Mohammad Fuad Jamal; مراد محمد فؤاد جمل
Wireless Sensor Networks (WSNs) are increasingly employed in various real-life applications, including smart cities, healthcare, and industrial systems; however, their limited computational resources make them more vulnerable to cyberattacks. Intrusion Detection Systems (IDS) that depend on Artificial Intelligence (AI) algorithms, including Deep Learning (DL) and Machine Learning (ML) algorithms, have shown their potential in detecting complex threat patterns. However, their high processing requirements restrict their deployment in WSN environments with constrained resources. This study solves the trade-off between detection accuracy and computational complexity. And it provides an energy-efficient deep learning model for intrusion detection in WSNs, in order to detect the latest cyber threats while maintaining high accuracy and low computational cost. We utilized the quantization optimization method over the Neural Network (NN) model to mitigate energy consumption. The findings of the study showed that the quantized model significantly reduces computational cost and memory usage while preserving good detection performance. The final prediction was made by employing an ensemble approach and combining all three models: DT, SVM, and the quantized NN. The final results showed that the ensemble model achieved an accuracy of 0.9238 with a prediction time of 4.1391 seconds, demonstrating a strong balance between detection effectiveness and computational efficiency. In addition, the ensemble attained high precision (0.9965), recall (0.9232), and F1-score (0.9585), indicating reliable detection of both normal and malicious traffic. Compared with individual models and evaluations on the NSL-KDD dataset, the proposed approach exhibited improved robustness and generalization while maintaining a compact memory footprint suitable for resource-constrained WSN environments.
An Advanced Deep Learning Framework for Real-Time Breast Cancer Lesion Analysis in Clinical Ultrasound
(Al-Quds University, 2025-05-20) Suliman Imad Suliman Thwib; سليمان عماد سليمان ذويب
Breast cancer represents the most common malignancy among Palestinian women, accounting for 32.1% of all female cancers with a concerning 52% of cases diagnosed at advanced stages. Early detection is hindered by multiple challenges, including limited access to specialized radiologists, inadequate digital infrastructure, and socioeconomic barriers. This research develops and validates an advanced deep learning framework for real-time breast cancer lesion analysis in clinical ultrasound videos, specifically designed to address these challenges within the Palestinian healthcare context. The research encompasses the entire analytical pipeline from contrast enhancement through detection to tracking. A comprehensive dataset of 17,903 ultrasound cases was curated from two Palestinian healthcare facilities, including 11,383 images from the Dunya Women's Cancer Center (2018-2023) and 6,520 video frames from Augusta Victoria Hospital (2024-2025). This locally representative dataset, including cases with confirmatory biopsy results, ensured the clinical relevance and cultural appropriateness of the developed system. The framework integrates three key technical innovations. First, a systematic evaluation of contrast enhancement techniques identified Contrast-Limited Adaptive Histogram Equalization (CLAHE) with a clip limit of 1 as optimal for breast ultrasound preprocessing, achieving superior Enhancement Measure Estimation (23.670) and Peak Signal-to-Noise Ratio (24.359) with minimal computational overhead (0.536ms). Second, comprehensive comparison of You Only Look Once (YOLO) architectures established YOLOv11-L as the most effective model for lesion detection, achieving mean Average Precision of 0.93 and sensitivity of 0.88 while maintaining real-time performance (4.6ms). Transfer learning further improved performance, with the TL12 approach achieving mAP of 0.955 and sensitivity of 0.938. Third, a novel hybrid Detection-Based Tracking (DBT) approach combining Kernelized Correlation Filter (KCF) tracking with YOLO verification demonstrated superior performance for lesion tracking, achieving success rates of 0.976 for benign and 0.984 for malignant lesions when combined with CLAHE preprocessing. Performance evaluation using ultrasound video sequences with confirmatory biopsy results validated the framework's clinical applicability. The combined pipeline maintained real-time processing capability (approximately 54 frames per second) while achieving high accuracy in lesion detection and tracking. These results demonstrate that deep learning approaches can effectively enhance breast ultrasound interpretation capabilities in resource-constrained healthcare environments. This research represents a significant contribution to addressing breast cancer detection challenges in Palestine by providing a computationally efficient, clinically applicable framework that complements limited specialized expertise. The methodology established for developing context-specific AI systems for medical imaging offers a blueprint for similar initiatives in other resource-constrained healthcare environments. Future work should focus on implementing multiple object tracking capabilities, extending to 3D ultrasound analysis, and conducting larger multi-center clinical validation studies.
Framework for Augmented Reality User Experience: Automated Measures of User Experience and Associated Factors
(Al-Quds University, 2025-01-12) Balqees Ahmad Khalil Awawdeh; بلقيس أحمد خليل عواوده
Background: Augmented reality (AR) technology is rapidly advancing, offering new possibilities for user experience (UX) across fields such as education, tourism, and architecture. Nevertheless, the development of automated frameworks that thoroughly evaluate UX is crucial, particularly in terms of learnability, efficiency, effectiveness, and memorability, based on established human-centered interaction standards. Objective: This study aims to develop a framework for testing user experience in AR applications, with a primary focus on evaluating usability and user engagement metrics. The framework uses established Human-Computer Interaction principles as a reference for automated measures, providing a comprehensive and objective assessment tool to support the development of intuitive, user-friendly AR interfaces that enhance user engagement. Method: The automated framework was applied to assess usability metrics across three AR applications—Farah App (educational), Dar Al Consul (tourism), and EasyApp (architectural)—involving a group of 20 users across different age groups. Data on user interactions were collected and analyzed, while Participant responses was gathered through questionnaires to complement the automated analysis. Results: Findings revealed notable differences in ease of learning, effectiveness, and memorability among the applications. Farah App showed particularly high learnability, especially among younger users, making it a promising educational tool for that demographic. Both Dar Al Consul and EasyApp demonstrated consistent usability across age groups, though minor improvements in task flow and error reduction could further enhance user efficiency. Conclusion: This study highlights the importance of using HCI-based frameworks to evaluate AR user experience. The developed framework provides comprehensive and objective measurements that support enhanced UX by offering data-driven insights for more intuitive design. The results indicate that AR developers should focus on age-appropriate interactions, streamlined task flows, and efficient feedback mechanisms to improve user engagement and ensure effective, sustainable experiences across a range of AR applications.
Comparison of Native and Cross-Platform Mobile Development Tools for Android Mobile Application
(Al-Quds University, 2024-08-21) khalid waleed izzat zohud; خالد وليد عزت زهد
The main objective of this study is to compare the different programming languages used in creating smartphone applications for Android phones in a native or cross-platform format. This study examines the performance of five different programming languages by creating an application that runs the same tasks in each programming language. The comparison was based on the amount of time spent on each executed task. This study uses an experimental approach to explore and discuss the approaches and applications of cross-platform application development. Therefore, a sample project (Test Bench) was implemented with a native framework android studio with Kotlin programming language, followed by four cross-platform frameworks visual code with React Native and Cordova framework, android studio with Flutterfarmworkand visual studio with Xamarin framework. The collection and recording of information and analysis were for efficiency, workload, software procedures, and performance by recording the time of tasking on the program that was created for this purpose, where the researcher recordedthe advantages and disadvantages of each framework, and eventually compare the five frameworks (native and four cross-platform languages). The results of this study shows that every cross-platform, such as Xamarin, has its own advantages and disadvantages. It has the best result for image capturing and compressed files, but it came to the end regarding the time in read/write storage. Xamarin is the best choice for c#.NET and desktop developers who want to start developing mobile applications. The result showed that React Native, Cordova is the best choice for web developers, Native frameworks like Kotlin and Flutter for desktop developer
A Framework for Integrating Technical Performance Metrics and SUS Testing to Enhance User Experience in Augmented Reality
(Al-Quds University, 2025-01-12) hadeel adnan atallah farash; هديل عدنان عطالله فراش
Evaluating the performance of augmented reality (AR) applications is essential for optimizing user experience, improving interaction quality, and advancing AR technology. Traditional AR performance studies typically focus on factors such as response speed, tracking accuracy, resource usage, and their impact on user experience. These studies aim to reduce energy consumption, enhance application efficiency, and enable the integration of emerging technologies like artificial intelligence (AI). Additionally, understanding the influence of AR applications on user interaction is crucial for developing applications that produce positive results. Identifying effective methods for content delivery through comprehensive performance comparisons is key to ensuring an optimal user experience. Despite significant research in the field, many studies lack an integrative approach that considers the full range of technical performance factors. This research introduces a framework designed to provide developers with a robust, automated performance evaluation tool that enables them to enhance the AR user experience. The developed tool automatically collects and analyzes user interaction data—such as gameplay data—and integrates this with user feedback obtained from post-session surveys. This dual-method approach helps identify performance weaknesses and their underlying causes, providing developers with targeted insights for improving the application's quality. The tool was tested across three AR applications in distinct categories—educational, tourism and cultural heritage, and gaming—demonstrating its effectiveness in delivering valuable, context-specific performance metrics. The most significant findings from the analysis underscore the importance of AR tracking accuracy in achieving high user satisfaction. Applications with high tracking accuracy, such as the Dar Al Consul App, demonstrated superior user engagement and satisfaction levels compared to applications with stable technical performance but less accurate tracking, such as the Farah App and EasyAPP. Additionally, the study found that while technical metrics like CPU, GPU, latency, and memory usage were consistent across user groups, user satisfaction was more strongly correlated with usability factors and interface intuitiveness, particularly for younger users who benefit from age-appropriate designs. In addition to its real-time analysis capabilities, the tool features a data storage function that supports long-term performance tracking, laying the foundation for AI-driven models in the future that can learn from the strengths and limitations of AR applications. By combining user feedback with AI, the framework offers a holistic approach to performance evaluation, assisting developers throughout both development and pre-release stages. This research bridges gaps in current AR evaluation methods, providing a comprehensive approach that encompasses the critical technical, psychological, and social aspects of AR application performance.

Browse

Recent Submissions