M.Phil / MS
Permanent URI for this collectionhttps://repository.cuilahore.edu.pk/handle/123456789/36
This collection archives the complete set of theses produced by students of the COMSATS University Islamabad, Lahore Campus.
Browse
12 results
Search Results
Item Securing Email Communications: Advanced Approaches to Detecting Phishing Through Spam(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Amna Asif; SP23-RCS-006; Dr. Atif Saeed; LHR TP 9484The growing complexity of email-based communication has made it difficult to correctly identify the spam emails which provide significant hazards such as phishing, malware distribution and data breaches. Conventional spam detection algorithms that use static rule-based techniques like Keyword-Based Filtering, Blacklist Filtering, or rely on individual machine learning algorithms like Naïve Bayes, Logistic Regression, K-Nearest Neighbors, usually fall behind in increasing false positive and decreasing accuracy. This study used an optimal ensemble-based approach for spam email identification using Gradient Boosting Machine and Extreme Gradient Boosting algorithms and aims to enhance spam email detection accuracy through efficient hyper parameter tuning of these machine learning algorithms. Primarily, baseline models were trained on default parameters and then performance of these models was improved through randomized search cross validation method to examine the tuning space of hyper-parameters for efficient hyper-parameter values. By using the Enron dataset, a publicly available extensive collection of actual email data containing 33639 labeled emails, models were assessed using key metrics such as accuracy, F1-score, recall, precision, and ROC-AUC. The experimental results of this study presented that the performance of both algorithms was enhanced by the hyper-parameter tuning when contrasted with the baseline models. The modified XGBoost model outperformed the baseline version and other competitive models with an accuracy of 98.65%. Furthermore, on tuned parameters GBM performed well, demonstrating the effectiveness of ensemble algorithm approach. The results highlight the superiority of ensemble algorithms for challenging classification problems and validate the significance of hyper parameters modification in improving model performance. This research offers useful insights for enhancing cybersecurity measures in email communication systems by developing strong spam detection frameworks.Item Semantic Precision: Forecasting Healthy Practices for Diabetic Patients Using Social Media Perspective and Medical Models(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Hina Idrees; SP23-RCS-015; Dr. Abid Sohail; LHR TP 9483Hyperglycemia mellitus is a chronic health condition that affects millions of people globally, posing significant challenges to both individuals and healthcare systems. The condition arises due to the body's inability to produce enough insulin, as seen in Type 1 diabetes, or its inability to effectively utilize insulin, as in Type 2 diabetes [1]. Managing diabetes requires a multifaceted approach, including regular monitoring of blood sugar levels, adherence to medication regimens, maintaining a balanced diet, engaging in physical activity, and being vigilant for potential symptoms [2]. Poor management can lead to severe complications such as cardiovascular diseases, kidney damage, nerve damage, and vision impairment [3]. This thesis proposes a novel framework to simplify and enhance diabetes care by leveraging real world data, process modeling, and advanced machine learning techniques.Item A Methodological Approach for Classifying and Differentiating Business Processes Using Tasks Label(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Sidra Anwar; SP23-RCS-027; Dr. Abid Sohail; LHR TP 9492It is crucial for two organizations to integrate their business process models when one acquires the other or merges with another organization. This research is motivated by the problem of matching processes with semantically similar activity names and similar process maps that make alignment difficult. Introducing a new approach based on the Domain Specification Mapper along with the current state-of-art transformer models, BERT for semantic augmentation and contextual interpretation of activity labels. The method combines semantic analysis, syntactic features and machine learning algorithms to show that DSM does not only increase the effectiveness of the transformer models such as BERT, but also increases the effectiveness of other machine learning models. Although traditional classifiers and semantic syntactic features failed to provide a correct classification of processes, DSM with BERT can successfully address these problems. Furthermore, the integration of DSM with other classifiers also showed good results and improved the classification rate. This approach facilitates integration of process, increases accuracy and aligns operations which makes it ideal for mergers and acquisitions.Item A Benchmark Corpus and Methods for the Urdu Qur’anic Question-Answering System(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Abida Ashraf; SP23-RCS-002; Dr. Rao Muhammad Adeel Nawab; LHR TP 9497Question Answering System aims to provide answers to a question posed in natural language. A question can be factoid and non-factoid. With the rapid increase in technologies, the QA task become one of the most challenging and important tasks of Natural Language Processing (NLP). In previous studies, the problem of the QA system for the Holy Quran has been explored for Arabic, English, and Indonesian languages. Urdu has more than 170 million speakers worldwide yet considered low low-resource language. However, the problem of the QA system for the Holy Quran has not been explored for the Urdu language due to the unavailability of the dataset. To overcome this research gap, this study developed a benchmark corpus of the Holy Quran in the Urdu language. We developed the corpus by carefully mapping the QRCD corpus which was originally in Arabic language into Urdu. For the mapping of answers, we take the expertise of data annotators who have a deep knowledge of the Holy Quran in both Arabic and Urdu language. We conducted experiments in a state-of-the-art BERT based LLMs. We evaluate the proposed techniques using Exact Match, partial Reciprocal Rank, and F1 score, and expert judgment.Item Towards Better USD/PKR Exchange Rate Forecasting Using Public Sentiment Analysis(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Muhammad Rizwan; CIIT/SP23-RCS-022/LHR; Dr. Abid Sohail; LHR TP 9995Exchange rate forecasting plays a crucial role in financial planning, investment decisions, and economic policy formulation. Traditional forecasting models primarily rely on historical financial data and macroeconomic indicators; however, these models often fail to capture the influence of public perception and market sentiment on currency movements. This study aims to improve the forecasting accuracy of the USD/PKR exchange rate by incorporating public sentiment analysis derived from online sources such as social media platforms, news articles, and financial forums. The research collects historical exchange rate data along with large-scale textual data reflecting public opinions related to economic conditions and currency fluctuations. Natural Language Processing (NLP) techniques are applied to preprocess the textual data and perform sentiment analysis, categorizing the information into positive, negative, and neutral sentiments. These sentiment indicators are then integrated with traditional time-series forecasting models and machine learning techniques to predict future movements of the USD/PKR exchange rate.Item Detecting Breast Cancer from Histopathology Images Using CNN-Based Framework(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Tahreem Zara; CIIT/SP23-RCS-028/LHR; Dr. Zulfiqar Habib; LHR TP 9706Breast cancer remains one of the most frequently diagnosed cancers among women globally, with histopathology analysis serving as the gold standard for diagnosis. But manual examination of histopathology slides takes a long time, is prone to making mistakes, and often has differences between observers, especially in complicated cases. To overcome these constraints, this study introduces a lightweight and deployable deep learning framework designed to automate breast cancer detection utilizing histopathology images from the BRACS dataset. The proposed framework utilizes the DenseNet169 model pre-trained on ImageNet; we fine-tuned the last 30 layers after testing different depths. We resized the input images to 1024×1024 to preserve important morphological details, and we employed simple augmentations, such as rotation and flipping, during training. Additionally, a semi-automated segmentation and annotation strategy was developed using k-means clustering based on RGB color and shape features, generating binary masks that were further validated by a histopathologist. Both classification and segmentation models were implemented in Keras on a CPU-based environment with 32 GB RAM. We evaluated our models using scikit-learn. The classification achieved 79% accuracy on the 3-class test set and 90% accuracy on the binary test set (normal vs. invasive carcinoma), demonstrating competitive results compared to state-of-the-art models. The evaluation metrics used for both models were precision, recall, and F1-score; classification performance was further assessed using accuracy, and segmentation performance was evaluated using the Dice coefficient. Our classification model (DenseNet169 with partial fine-tuning) has approximately 4 million trainable parameters, while the segmentation model (U-Net with ResNet34 encoder) consists of around 25 million parameters—both models being compact enough for deployment on standard desktop CPUs. The segmentation masks were generated using a novel semi-automated pipeline that balances visual coherence (shape and structure) with RGB clustering, allowing more precise annotation than conventional thresholding or fully automated tools. These findings highlight that a selectively fine-tuned CNN can deliver reliable diagnostic assistance, even in resource- constrained environments.Item Congestion Aware Bandwidth Optimization for Secure Healthcare Data Transmission Using Edge Computing(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Aiza Ijaz; CIIT/SP23-RCS-003/LHR; Dr. Tahir Maqsood; LHR TP 9700The increasing dependence on telemedicine and remote patient monitoring between IoT wearable medical devices has revolutionized modern healthcare services. However, managing security, privacy and effective communication of huge amounts of sensitive healthcare data remains a critical challenge. Existing blockchain systems, still promising and guaranteeing data integrity and decentralized control, clash with scalability, transaction delays and network traffic, especially when processing real time medical IoT data. These limitations block the timely delivery of critical healthcare information, probably affecting patient outcomes. This thesis presents the development of a blockchain telehealth system combined with Software Defined Networking (SDN) to reduce these challenges. The proposed framework employed the Ethereum blockchain smart contract for secure, decentralized authorization and access control of healthcare data, integrated with edge computing for temporary off-ledger data storage to relieve blockchain scalability issues. Furthermore, SDN is employed to allocate bandwidth, ensuring low latency transmission of critical patient data, especially during emergencies. The primary objectives of this research involve guaranteeing the security and privacy of healthcare data. Using edge computing improves storage and transmission efficiency and optimizes bandwidth allocation using SDN techniques. The proposed system introduces a novel solution by integrating SDN and blockchain. Experimental results show that the framework visibly improves bandwidth utilization, reduces latency for emergencies and increases overall system efficiency compared to old blockchain approaches. This research bridged the gap between data security and transmission efficiency in healthcare IoT systems. Future work will focus on integrating AI driven routing algorithms and energy efficient IoT models to further optimize performance and support large scale healthcare deployments.Item Secure and Reliable Medical Data Offloading in Mobile Edge Computing Leveraging Software Defined Networking and Blockchain(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Asmara Maryam; CIIT/SP23-RCS-009/LHR; Dr. Tahir Maqsood; LHR TP 9702The volume of medical data is growing rapidly due to the increasing popularity of the Internet of Things and other sophisticated medical devices, making medical data processing more and more difficult. As healthcare systems become increasingly data- driven, the need for secure and real-time access to patient information has grown significantly particularly during emergencies and critical procedures. However, the limited processing capacity of traditional hospital infrastructure makes it challenging to manage and transmit the large volumes of data generated by IoT-based medical devices. In this work, we present a framework that securely and reliably offloads patient data using a combination of Software-Defined Networking (SDN), Blockchain, and Mobile Edge Computing (MEC) technologies. Using mobile edge computing, cloud computing processing capacity can be distributed and mapped closer to users at the network edge, allowing healthcare systems to offload data efficiently. The proposed system enables healthcare data to be offloaded to nearby edge nodes with minimal delay, while ensuring that only authorized individuals such as doctors with patient consent can access or modify the information. SDN intelligently selects the best network paths based on real-time bandwidth and latency, while smart contracts deployed on the Ethereum blockchain manage access control, consent verification, emergency overrides, and data logging. However, ensuring reliability of offloading is critical because failure to process medical data may have catastrophic effects. Similarly, ensuring privacy and security of medical data is another challenge that needs to be addressed. We developed a three-layer architecture involving IoT devices, edge nodes, and the blockchain, and implemented modular smart contracts for various healthcare tasks, including usage tracking and ML model validation. Performance was evaluated using NS-3 for SDN simulation and Remix IDE for smart contract testing. Results show that our approach improved reliability, achieving a 96.5% packet delivery ratio, a 99.8% data integrity (hash match) rate, and a 24% reduction in average latency compared to traditional models. This demonstrates the potential of SDN and blockchain in making healthcare data offloading both secure and efficient.Item Intelligent Hotel Recommendation System Using Sentiment Classification and Machine Learning(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Muhammad Ahmad Yousaf; CIIT/SP23-RCS-019/LHR; Dr. Atif Saeed; LHR TP 9704Travellers are depending more and more on digital platforms to select appropriate lodgings as their dependence on online evaluations grows. However, current hotel recommendation systems sometimes offer general recommendations without being able to comprehend user-specific requirements or evaluate reviews according to specific service elements like location, cleanliness, food quality, and security. Additionally, these systems frequently ignore neutral thoughts, which lowers the recommendations' accuracy and personalization. This study's goal is to create an intelligent hotel recommendation system that combines cutting- edge machine learning and sentiment analysis methods to provide tailored, aspect-based recommendations. To achieve robust classification, the suggested model combines Random Forest with BERT (Bidirectional Encoder Representations from Transformers) for deep contextual sentiment interpretation. To handle a sizable dataset of hotel reviews gathered from Booking.com, the system uses natural language processing techniques including lemmatization, tokenization, stop-word removal, and feature extraction using TF-IDF. By classifying and analyzing reviews based on different hotel features, consumers may do query-based filtering. For instance, they might request hotels with high ratings for cleanliness or food quality. Common issues with current systems, such as cold- start issues, a lack of aspect-level insights, and inadequate user personalization, are addressed by this hybrid paradigm. The method improves the accuracy and applicability of hotel suggestions by precisely reading user preferences and attitudes. The study shows how deep learning and ensemble techniques may be used to create recommendation systems that are more context-aware and user- centric, which enhances decision- making in actual travel situationsItem Neural Network-Based Anomaly Detection in Cryptocurrency Transactions(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Rashida Noreen; CIIT/SP23-RCS-025/LHR; Dr. Tariq Umer; LHR TP 9705Cryptocurrency, especially the rapid growth in adopting bitcoins has led to a significant increase in fraud and illegal transactions, which faces serious challenges for financial integrity and digital trust. Methods of detecting traditional fraud are often reduced to detecting sophisticated patterns within the blockchain network due to decentralized and dynamic nature of transaction flows. This thesis presents a graph-based deep learning approach to detect discrepancy in cryptocurrency transactions, which uses a large, label dataset of bitcoin transactions collected over time. The functioning begins with the collection and prepressing of the elliptical dataset, which includes more than 230,000 edges representing more than 200,000 transactions and money flows represented as nodes. Dataset consists of three classes: license, illegal and unknown transactions. To extract meaningful insight from this data, a broad time-based and network-based feature engineering was performed. After feature engineering, the dataset was converted into a graph structure suitable for processing by the graph neural network (GNNS). Four state -of -the -art models trained and evaluated: Graph Attention Network (GAT), GATV2, Siamese GNN, Relational Graph Convolution Network (RGCN), and graph transformers. The model was trained and validated using stratified partitions to ensure balanced class representation. In the models, the Siamese GNN achieved the highest performance with an accuracy of 95.67% GAT followed closely with 94.90% accuracy, while graph transformers 94.60% and GATV2 achieved 94.46% competitive but slightly less results. RGCN showed 93.77% the least effective performance, especially struggling with unknown transactions classification. The results strongly suggest that graphs can play an important role in increasing the accuracy and reliability of the system that detects fraud in the blockchain environment, especially in accordance with dynamic and asymmetrical graphs.