http://dlib.iit.ac.lk/xmlui/handle/123456789/2761 2026-04-28T14:26:47Z 2026-04-28T14:26:47Z Kodithuwakku Arachchige, Yathindra http://dlib.iit.ac.lk/xmlui/handle/123456789/2799 2025-07-01T03:31:39Z 2024-01-01T00:00:00Z

DAO Compass: Risk Analysis of Proposals in Decentralized Autonomous Organizations for Informed Voting Kodithuwakku Arachchige, Yathindra "Decentralized Autonomous Organizations (DAOs) are pivotal in the blockchain ecosystem, fundamentally altering how decisions are made and executed within these decentralized entities. However, the integrity and security of DAOs can be compromised by malicious proposals that exploit vulnerabilities within the system. This research addresses a significant gap by introducing a novel risk analysis architecture aimed at identifying threats posed by abnormal proposals in DAOs. The architecture employs parametric analysis and address screening. Parametric analysis utilizes statistical outlier detection techniques to analyze proposal actions and detect abnormalities in function parameters based on historical data. Address screening assesses the integrity of associated addresses by cross-referencing them against known scam and sanctioned addresses, among other criteria. This dual approach enhances the detection of potential threats and improves the informed decision-making capabilities of DAO participants. By focusing on the specificities of smart contract interactions within DAOs, this solution extends beyond existing network-level security measures, contributing significantly to blockchain security and software engineering. To test the novel statistical component of the solution architecture, synthetic data generation was followed by evaluation metrics calculation, yielding an precision of 92.3%, accuracy 85%, f1-score 80% and recall of 70.5%. Based on the outlined benchmarking results, it is clear that the proposed novel algorithm performs comparatively better than the considered mainstream statistical algorithms. This research ultimately aims to fortify the reliability and trustworthiness of DAO operations, ensuring their role as secure platforms for collective decision-making."

2024-01-01T00:00:00Z Jayasuriya, Yasasa http://dlib.iit.ac.lk/xmlui/handle/123456789/2798 2025-07-01T03:28:58Z 2024-01-01T00:00:00Z

Monitoring the Mental Health of Employees Using Emotion Detection Jayasuriya, Yasasa In today's workplace, mental health is increasingly seen as a critical component of employee well-being, with clear implications for productivity, job satisfaction, and overall organizational climate. Traditional means of monitoring employee mental health, like as surveys and personal evaluations, are sometimes subject to biases and mistakes, resulting in delayed or inefficient assistance for those in need. Our research provides a cutting-edge desktop application meant to monitor and support employee mental health through real-time emotion recognition, employing powerful CNN and ADT. ADT is used to improve the model's robustness against adversarial examples, ensuring system reliability even in the presence of potentially deceptive inputs. By training the deep learning model on a large dataset of facial expressions, the application can accurately recognize a wide range of emotions, including happiness, sadness, stress, and anger, providing significant insights into employees' mental well-being. Comprehensive testing and user acceptance studies have confirmed that the application is highly effective in detecting mental health disorders early, allowing for timely and suitable therapies. The application was tested using the current CNN model which showed an accuracy of 73% when the testing area was complex with many people moving around. However the application’s accuracy improved up to 86.67% when tested in a more favorable environment with 150 participants. It is shown that the application's emotion recognition skills are highly accurate and trustworthy due to advanced picture processing and data augmentation methods. This study offered a cutting-edge method for monitoring employee mental health through automated emotion recognition using deep learning algorithms such as CNN. The initiative, which was recognised as an important invention in this sector, created a sophisticated online system that efficiently analyses emotional expressions, paving the way for automated mental health monitoring.

2024-01-01T00:00:00Z Mahil, Tuwan Shihan Rahim http://dlib.iit.ac.lk/xmlui/handle/123456789/2797 2025-07-01T03:25:56Z 2024-01-01T00:00:00Z

D-Forger: Low-Quality Neural Texture Video Forgery Detection Through Deep Learning Mahil, Tuwan Shihan Rahim "The detection of low-quality video forgeries, particularly those based on neural textures, is an increasingly rising challenge in digital media forensics. This thesis proposed D-Forger, a novel deep learning-based system designed to accurately identify these types of forgeries. Existing methods primarily target high-quality manipulations, often neglecting the subtler inconsistencies present in low-quality videos. This research addresses this gap by developing and optimizing feature extraction techniques and classification algorithms that are sensitive to the unique artifacts of low-quality forgeries. The proposed system's architecture and methodologies aim to enhance detection accuracy across varying video qualities, emphasizing neural texture anomalies that are frequently overlooked by conventional approaches. To achieve this, the study conducted a comprehensive review of current forgery detection techniques, identifying significant research gaps, especially in the domain of low-quality video manipulations. Extensive requirement gathering and system design phases were followed by the development of a prototype system. The implementation leveraged the FaceForensics++ dataset, focusing specifically on Neural Textures-based manipulations. The prototype's performance was evaluated and provided promising results. The D-Forger system achieved an accuracy of 96%, a precision of 0.97 for fake videos and 0.96 for real videos, and recall rates of 0.96 and 0.97 for fake and real videos, respectively, resulting in a balanced F1-score of 0.96 for both classes. Furthermore, the model demonstrated a robust ability to distinguish between fake and real videos, as evidenced by an AUC of 0.96. These results underscore the system's potential to effectively identify low-quality forgeries, marking a significant improvement over existing methods. The contributions of this research offer a novel approach to detecting neural texture-based low-quality video forgeries and provide a robust system architecture adaptable for low-quality video analyses. The findings highlight the critical need for specialized detection tools to mitigate the risks posed by deepfakes on social media platforms, where low-quality forgeries are more prevalent, with future work focusing on refining the detection algorithms and expanding the dataset to enhance the system's robustness and accuracy further."

2024-01-01T00:00:00Z Dharmavijaya, Thilanka http://dlib.iit.ac.lk/xmlui/handle/123456789/2796 2025-07-01T03:23:36Z 2024-01-01T00:00:00Z

Machine Learning Based Fraud Detection In Health Insurance Dharmavijaya, Thilanka "Insurance fraud is a growing problem in the economic sector, distinguished by its constantly changing nature and growing complexity. As fraudsters continue to develop new strategies, conventional methods of detection are proving insufficient, requiring the implementation of more efficient and creative alternatives. This study enhances the precision and adaptability of insurance fraud detection by applying an approach based on machine learning. The study suggests a new model that takes advantage of the natural data imbalance in fraud detection scenarios. This model not only improves the rates of identifying fraud but also offers valuable insights into how fraud processes work. This study aims to address the current limitations in technique and make a significant theoretical and practical contribution to the field of fraud prevention by combining domain-specific knowledge and utilizing advanced algorithms. The main objective of this study is to employ machine learning techniques for the purpose of detecting fraudulent activities in the insurance industry. At first, an Exploratory Data Analysis (EDA) was performed to get insight into the properties of the dataset. In order to tackle the issue of imbalanced data, where the number of fraudulent claims is substantially lower than legitimate ones, oversampling techniques were utilized to properly balance the dataset. After performing data preprocessing, process included transforming categorical variables into one-hot encoding. Additionally, the dataset was split into distinct training and testing sets. Subsequently, a range of machine learning techniques were employed, such as Random Forest, Decision Tree, and Support Vector Machine (SVM). Out of all them, the highest level of accuracy achieved was 0.87. The study improved models' performance by fine-tuning hyperparameters, addressing data imbalance and reducing fraudulent claims compared to authentic ones."

2024-01-01T00:00:00Z