UTAR Institutional Repository (Universiti Tunku Abdul Rahman)
Not a member yet
6132 research outputs found
Sort by
Enhancing deepfake detection generalization through component-based development in a web platform
No full textThis report outlines the design, development, and evaluation of a deepfake detection system aimed at providing an accessible and scalable solution for detecting manipulated media. The system leverages advanced machine learning models, including both single-model and ensemble-based detection methods, to identify deepfakes in images. The platform supports easy image uploads, efficient model processing, and reliable result presentation, offering users the ability to choose between various detection models based on their needs.
Key features include user authentication and role management, image validation, preprocessing, and real-time inference with confidence scores. The system utilizes a modular architecture to integrate new models seamlessly, ensuring scalability and maintainability. Performance benchmarks are met, including a processing time of less than 800ms per image and a 99.9% uptime for system reliability. The accuracy of the ensemble detection method is validated through extensive testing on benchmark datasets, achieving a high F1-score.
This project addresses the growing concern of deepfake threats in digital media and aims to provide an easy-to-use, robust tool for both non-technical users and advanced administrators. The system is designed with a focus on usability, accuracy, performance, and security, ensuring it meets the challenges posed by modern deepfake detection.
Keywords: Machine Learning, Generative AI, Deepfake, Component-based, Image Classification, Artificial Intelligence, AI Generalization
Subject Area: QA75.5-76.95 Computer Scienc
“When languages collide” A study on cross-linguistic phonological interference in multilingual communities in Malaysia
Get PDFThis study analyses the types of segmental and suprasegmental interference as well as the effect of executive functioning on the prevalence of phonological interference in Malaysian bilinguals and multilinguals. By focusing on four different language groups present in Malaysia, this study aims to identify the correlation between the inhibition of linguistic interference in the reading tasks and the inhibition of irrelevant stimuli. The segmental variants were examined by utilizing Weinreich (1953)’s framework while suprasegmental interference focused solely on stress placement and its role in polysyllabic words, compound words and at sentence-level. Variants produced in both sections were related to influence from the speakers’ dominant language and Selinker’s Interlanguage Theory (1972). Next, the role of executive functions in inhibitory control was investigated via the results of the Simon Task and Stroop Task to test the bilingual advantage hypothesis as well as whether better performance in executive function tasks translates to a lower susceptibility to phonological interference. The data of this study revealed that the Mandarin Chinese-dominant speakers produced the greatest number of variants overall while the English-dominant speakers and Tamil-dominant speakers showed the least variants in segmental interference and suprasegmental interference respectively. Despite significant anomalies in phonological variants produced and performance in Simon and Stroop Task, one notable correlation between low variant production and stronger performance in the Stroop Task was identified in English-dominant speakers, suggesting that language dominance plays a role in inhibition of interference and irrelevant stimuli
The adsorptive removal of antibiotic ciprofloxacin from aqueous medium by using banana peel
Get PDFThis research explores the adsorption of ciprofloxacin (CIP) from aqueous solutions using banana peel (BP) as an adsorbent. The goals include figuring out the removal efficiency under various conditions, characterizing BP and figuring out the isotherms and adsorption kinetics. According to the study, CIP's adsorption onto BP rose with increasing agitation rate, adsorbent dosage, and contact time. The adsorption process adhered to the Freundlich isotherm and followed the Pseudo-Second Order kinetic model, showing optimal effectiveness at pH levels above 4. The maximum CIP adsorption capacity by BP was 37.037 mg/g. SEM and AFM analyses verified surface changes in BP before and after adsorption, while the FTIR study suggested functional group involvement in the CIP-BP interaction. The significant factors that mainly contributes to the adsorption process was determined by Plackett-Burman Design, and the optimized condition equation was determined by Response Surface Methodology. BP shows great potential as an eco-friendly and cost-effective adsorbent for the removal of CIP from water, and additional modifications could further enhance both its adsorption capacity and practical utility
Physicochemical and antioxidant profiling of fermented milk drinks
Get PDFFermented dairy-based beverages have been consumed for centuries and continue to be widely popular in contemporary markets. Numerous brands of fermented milk drinks are commercially available, yet their physicochemical and antioxidant properties vary depending on the production process. This study aimed to analyse the physicochemical and antioxidant properties of selected commercially available natural flavoured fermented milk beverages produced by different food companies. The samples included three cultured milk drinks (Yakult®, Vitagen®, and Betagen) and three yogurt drinks (Farm Fresh®, Lactel, and Yobick). Physicochemical parameters, including pH, total titratable acidity (TTA), total soluble solids (TSS), and colour, were assessed, along with antioxidant properties, including total phenolic content (TPC), total flavonoid content (TFC), ferric reducing antioxidant power (FRAP), and DPPH radical scavenging activity (RSA). Data were analysed using nested analysis of variance (ANOVA) followed by Tukey's HSD test to determine significant differences (p≤0.05). Results indicated that Yakult® cultured milk exhibited the lowest pH (3.98±0.044) and the highest TTA (6.79±0.91), reflecting the highest acidity and a sour taste profile. Betagen cultured milk had the highest °Brix value (17.4±0.10), suggesting greatest degree of sweetness. Cultured milk drinks generally appeared darker and more yellowish, whereas yogurt drinks tended to be lighter and exhibited less yellow intensity. In terms of antioxidant properties, Farm Fresh® yogurt drink had the highest TPC (1.084±0.006) and TFC (5.148±0.071), indicating greatest phenolic and flavonoid content due to the presence of a greater number of probiotic strains. Lactel yogurt drink exhibited the highest FRAP (3012.50±12.37) and RSA (50.21±0.45), suggesting superior antioxidant activity attributed to its high protein content in fermented milk base. Fermentation conditions characterized by moderate acidity, the inclusion of acidproducing and multiple probiotic strains, and the use of a high-protein milk base were essential in order to produce fermented milk beverages with superior
antioxidant quality
Mobile assets monitoring using RFID for UTAR Hospital
Get PDFMost of the service sectors like hospitality, education, banking, and other service sectors were manually entering the data of assets into the database. Sometimes, it causes data errors and inaccuracies since there are human typing errors or overlooking the number of lines of assets. Recently, AI and IoT are the most convenient technologies and are very useful in all sectors, especially in the service sector. They help sectors produce the most profit and productivity through technology. Since RFID is more convenient to use to check and update the data of assets, it is suitable for use in an asset monitoring system.
The outcome of the project is a mobile application for the asset monitoring system using RFID for UTAR Hospital. The purpose of developing this mobile application is to keep track of the real-time incoming assets of the hospital. The assets will be labelled with RFID tags and registered using a mobile RFID reader, then the information of RFID tags will be uploaded to the server. In case of finding assets needed by hospital staff, they can immediately find the assets through the server and confirm the location of the asset
Car rental using mobile application development
Get PDFThis project aims to develop an innovative mobile car rental system that focuses on modernizing the traditional car rental process by integrating real-time tracking features, comprehensive customer support, and an intuitive user interface (UI). The transformation brought by mobile technology to the car rental industry has completely changed the way users interact with rental services. By analyzing existing car rental mobile applications such as GoCar Malaysia, TREVO, and Moovby, this study identifies strengths and weaknesses to inform the development of the proposed system. To address the problems found in these existing systems, the project objectives for the proposed system are discussed. The advantages of the existing systems have been adopted and integrated into the proposed system to enhance its functionality. In the proposed car rental mobile application system, it provides the modules included profile module, booking module, history module, and tracking module. The project uses the Android Studio and Flutter programming languages to develop a mobile-based car rental mobile application system. The choice of a mobile application is driven by the fact that nearly everyone uses mobile devices and to transform the traditional car rental booking process into a modern, more convenient experience. Through this project, our goal is to transform the car rental experience and provide users with a seamless, efficient, and modern rental service platform. This effort not only solves current industry challenges but also sets new standards for convenience, reliability, and customer satisfaction in the car rental sector
Empowering students in network security : A web-based learning strategy with the arcs model
Get PDFAccording to nowadays, technology has become more advanced in the world. In Malaysia it also becomes the hot topic between the society. With the advancement of technology, people find out that technology is a double-edged sword to our society. Especially for students, they need to know how to protect themselves in network. It brings out that network security is a critical skill for students nowadays. However, the traditional methods of teaching network security lack engaging content and are expensive. This project proposes a website which uses the ARCS model that can teach students network security concept in a more efficient and affordable way. The website will include PowerPoint slides, quizzes, video about hands-on labs and tutorials that will help students learn the essential skills they need to protect themselves online and gain new knowledge. Not only that, but the website will also have some mini games which will further enhance engagement and reinforce key concepts for students when they are playing it. The website also can help the students to reduce learning time. In summary, this project is to enhance the student to be aware of the problem about network security. Students also can learn how to protect their personal privacy when using the network and let the student know how they can solve it if they face the network security problem
Hotel feedback system
Get PDFThis project is a hotel feedback system application in Malaysia it is for academic purpose , focusing on Natural Language Processing (NLP) to enhance customer satisfaction and customer experience.
The current system has several issues, such as requiring customers to spend a long time providing feedback, inefficient data analysis of feedback, and hindering hotel management from understanding customer emotions effectively. The purpose of this project is to develop a hotel feedback system that simplifies the feedback process for customers while enabling hotel management to easily assess their real experiences and sentiments. To address these problems, the system implements speech-to-text functionality, a categorization feature, and sentiment analysis.
The method used in this project is prototyping, which allows for quick refinement of requirements and system development. The expected outcome is a system that enables users to provide feedback quickly, processes feedback data efficiently, and helps hotel management better understand customer sentiments. Additional planned features include voice recognition, feedback categorization, and sentiment analysis to interpret customer emotions.
This feedback system application will undergo testing, and based on user feedback, necessary improvements—such as adding missing features or fixing functional errors—will be implemented by the developer
Flood frequency analysis between annual maximum series (AM) and peaks-over-threshold (POT) series in Selangor, Malaysia
Get PDFFlood Frequency Analysis (FFA) refers to a statistical approach for estimating the
probability and magnitude of flood occurrences. Annual Maximum (AM) is the most applied approaches in FFA, it focuses on the most extreme event during the period of time. In fact, it is also crucial to take both smaller and frequent flood events into consideration when performing flood frequency analysis. Therefore, another approach named Peaks-Over-Threshold (POT) method provides a more precise way of computing occurrences of floods by including noteworthy high flow events (even if they are not the most extreme of the year), but it is usually underemployed because of its complexities in threshold selection. This research attempts to compare these two approaches in the flood frequency analysis inb Selangor. This study employs the L-moment approach to estimate the parameters of three candidate distributions, which are the generalised Pareto (GPA) distribution, the generalised extreme value (GEV) distribution, and the generalised logistic (GLO) distribution. Then, the L-moment diagram will be implemented to ascertain the optimum distribution for the data series. Additionally, each data series' distribution performance will be evaluated using the goodness-of-fit test and efficiency assessments, namely mean absolute error (MAE), root mean square error (RMSE) and BIAS. This study aims to determine whether the AM or POT technique associated with one of the three distributions above yields a more dependable and
accurate estimate of flood frequency in Selangor. This study can improve the way
that flood risk is assessed and managed. It also might help in building infrastructure and flood control solutions, especially for flood-prone regions. Based on the analysis conducted across 13 streamflow stations in Selangor, the POT approach was found to outperform AM in 10 stations, indicating its effectiveness in capturing a broader range of flood events. Among the three candidate distributions, the GPA distribution is being selected at 9 out of 13 stations, particularly due to its lower MAE, RMSE, and Bias values. For instance, POT-GPA combinations at stations like Sg. Bernam At Jam. Skc, Selangor and Sg. Selangor at Rantau Panjang recorded significantly lower MAE, RMSE, and Bias values compared to GLO and GEV. GLO and GEV was selected at 2 stations at 1 station respectively, indicatin comparatively less consistent performance. These findings suggest that the POT and GPA combination provides a more reliable and accurate estimate of flood frequency in Selangor. The results of this study are to support improved flood risk assessment and infrastructure planning in Selangor and similar flood-prone areas
Enhancing UTAR freshman mental health screening: integrating power automate to improve counselling support
Get PDFDepression may be the second most common global disease burden after HIV/AIDS in the year 2030. Approximately one million people in Malaysia aged 16 years and above suffer from depression. In alignment with ongoing efforts to promote mental health support, UTAR counsellors conducted mental health screening tests for freshmen using the Warwick-Edinburgh Mental Well-being Scale (WEMWBS) during orientation week. The WEMWBS test is designed to measure the mental well-being of individuals. However, the current screening process lacks immediate feedback or actionable suggestions for UTAR students. Additionally, there is uncertainty regarding the construct validity of the WEMWBS questionnaire in capturing the underlying psychological structure. Therefore, this research proposes an automated support system using Microsoft Power Automate to enhance UTAR freshman mental health screening procedures. The workflow may reduce the student's waiting period for support and the counsellors' workload by streamlining the management of student responses. Moreover, a series of multivariate analyses was conducted to gain deeper insight into the underlying structure of the well-being data. Particularly, Principal Component Analysis (PCA), Exploratory Factor Analysis (EFA), Confirmatory Factor Analysis (CFA), and Structural Equation Modelling (SEM) were applied to identify and validate the core dimensions of student mental well-being. A series of reliability analyses and convergent and discriminant validity were tested throughout the analysis. In conclusion, this research successfully achieved all the objectives by developing a working automation system and identifying key factors to student mental well-being. The EFA identified three underlying factors, namely Positive Emotion, Personal Growth, and Social Exploration. Among the three identified factors, Positive Emotion was the most influential. Promoting emotional positivity may most effectively enhance student well-being, offering a practical approach to improving university mental health support