Ilorin Journal of Computer Science and Information Technology

ENSEMBLE MODEL FOR THE DETECTION OF PHISHING URLs

Aboho Demenongo, Agaji Iorshase — Sun, 30 Jun 2024 00:00:00 +0100

Context: The rise in internet usage has led to a surge in online security risks, with phishing attacks being a major cause for concern. These attacks frequently involve the use of misleading web addresses to deceive users into revealing sensitive information. Detecting such harmful URLs is vital for safeguarding user data and online privacy. Objective: In this work, we present a combined model for identifying phishing URLs, harnessing the capabilities of machine learning algorithms, particularly Naive Bayes and Random Forest. Method: Our approach began by collecting a diverse dataset of web addresses, encompassing both legitimate and phishing URLs. We then performed feature extraction to represent these web addresses as numerical vectors, capturing their lexical features. These numerical features serve as input for training two well-known machine learning models: Naive Bayes and Random Forest. The Naive Bayes classifier, rooted in probabilistic principles, offers simplicity and efficiency in classification tasks. It constructs models based on the conditional probabilities of features given the class label, enabling it to make informed predictions. On the other hand, Random Forest, an ensemble learning technique, combines predictions from numerous decision trees to enhance accuracy and robustness. To make a final prediction, we employed a voting classifier that incorporates the predictions of Naive Bayes and Random Forest. Results: The result of our experiment indicated the efficacy of the voting classifier in identifying phishing URLs, as the model achieved an accuracy rate of 94.10%. We evaluated the performance of the voting classifier using various metrics, such as accuracy, precision, recall, and F1-score, revealing the ability to effectively differentiate between legitimate and phishing web addresses. Conclusions: This work contributes to the advancement of phishing URL detection methods, providing insights into the suitability of Naive Bayes and Random Forest for this purpose. Moreover, the proposed approach can aid in the development of more robust and accurate phishing detection systems, thereby enhancing online security and protecting users from cyber threats.

The Development of An Enhanced Voice Architecture Using Convolutional Neural Network (CNN)

Ibiyinka Temilola Ayorinde, Babatunde Alexander Abiola — Sun, 30 Jun 2024 00:00:00 +0100

Speech recognition and synthesis technology offers a natural interaction method for numerous computing tasks. It permits users to communicate with computers naturally using spoken language, requiring very little training. The Voce architecture is designed to be simple and it consists of a speech synthesizer called the FreeTTS. It also has a speech recognizer, called Sphinx4, which continuously listens for incoming audio data from the user’s audio hardware. Both Sphinx4 and FreeTTS libraries are difficult to implement, offering complex application programming interface (API). This leads to the major shortcoming of the Voce architecture since it only supports Java and C++. Also, the Voce architecture can only handle speech synthesis and recognition. Hence, this study builds a python language support for the Voce library as well as implement improvements to the library by including audio source separation and classification to the base library. The strengths of the voice synthesis and recognition libraries available in the Voce architecture were harnessed instead of building a new one from scratch. A wrapper was created for use in the python programming language using Py4j. Audio source separation was implemented using nussl, a flexible, object oriented Python audio source separation library while audio source classification was implemented in Keras using a Convolutional Neural Network. The network was trained using a suitable corpus and used to classify supplied audio input. The extended Voce library built in this study provides application developers with an Open Source, cross-platform library for speech synthesis, speech recognition, audio source classification and separation support with a simple API that can be used in python programming language. This new library will further promote the use of speech interaction technology by application developers.

Leveraging Machine Learning in Classifying Fraudulent and Legitimate Transactions in Banking Sector

Monday Jubrin Abdullahi — Sun, 30 Jun 2024 00:00:00 +0100

ABSTRACT:

This research explores the application of machine learning, specifically the Artificial Neural Network (ANN) algorithm, for the identification of fraudulent and legitimate transactions in the context of a European transaction dataset. The investigation encompasses various phases, including data preprocessing, exploratory data analysis, and the implementation of theSynthetic Minority Oversampling Technique(SMOTE) Tomek technique to address the imbalanced nature of the dataset. Leveraging Python libraries such as pandas, matplotlib, seaborn, and scikit-learn, the study incorporates advanced techniques like Principal Component Analysis (PCA) for dimensionality reduction. The ANN algorithm is chosen for its adaptability to complex patterns and high-dimensional datasets, and a graphical user interface (GUI) using the tkinter library to enhance user interactions with the developed sentiment analysis model. The results of the study demonstrate the efficacy of the proposed sentiment analysis model, with the ANN classifier achieving an impressive accuracy of 97.86%. This outperforms existing studies, such as Asha and Kumar's use of Support Vector Machine (SVM) (85%), K-Nearest Neighbour (KNN) (79%), and ANN (89%), and Yee, Sagadevan, and Malim's Bayesian network classifiers (95%). The model excels in minimizing false positives and false negatives, as indicated by maximized precision, recall, and F1-score metrics for both classes. The confusion matrix reinforces the model's outstanding performance, showcasing its potential for real-world applications in fraud detection within the banking sector. The Study's significance lies in its combination of sophisticated techniques, including feature engineering, algorithm selection, and a user-friendly interface, contributing to advancements in transaction classification and fraud detection in the financial domain.

Keywords: Machine Learning, Fraud Detection, Artificial Neural Network, Imbalanced Dataset, Graphical User Interface

Performance Evaluation of Some Selected Image Encryption Algorithms

Sun, 30 Jun 2024 00:00:00 +0100

Context: A medical image contains sensitive information critical to patient’s diagnosis by medical personnel hence, requires privacy and optimum security to preserve for future use. Encryption algorithms are known to be used to secure medical data such as images but their performance varies. Most encryption algorithms are characterised by high computational time, high memory usage and image lossy problem after encryption. Objective: This research addresses some of these problems by comparing the performance of four encryption algorithms: Advanced Encryption Standard (AES), CAST, Blowfish, and Enhanced Data Encryption Standard (E-DES) algorithms when applied on medical images. Method: The evaluation was conducted using five image samples of varying sizes, and several metrics were observed, including encryption time, decryption time, output byte, memory usage, mean square error (MSE), and peak signal-to-noise ratio (PSNR). Results: The experimental findings indicated that E-DES demonstrated superior performance in encryption time compared to other encryption algorithms, with an execution time of 4.08, 3.93, 3.16, 3.89 and 4.00seconds for Image A, Image B, Image C, Image D and Image E respectively while AES has the lowest memory usage. Conclusions: The findings of the comparison has added to the understanding of the effectiveness of specific image encryption algorithms and which can assist in resolving image security issues more quickly.

Design and Implementation of a Client-Server Model for Campus Office Chat Systems

Augustine OBAYUWANA — Sun, 30 Jun 2024 00:00:00 +0100

In today's business environment, many employees work remotely, and chat applications can be a critical tool for maintaining a sense of connection and community within a team. They can also help to ensure that everyone is on the same page, even when working from different locations. Chat applications help to increase efficiency by reducing the time and effort required to complete tasks. With quick and easy communication and collaboration, team members can work more effectively, reducing the risk of errors or misunderstandings. This project focuses on solving the problem of remote office communications by developing a campus office chat application for Realtime communication in Nigerian universities using the University of Benin as a case study. The client-server model chat system consists of three layers, presentation layer or client-side which is the front-end part of the application the user interacts with built with react.js a JavaScript frontend framework, the logic layer or server-side which contains the backend logic of the chat application built with Node.js a backend JavaScript Framework and, the data tier which stores several states of the user’s information and data. Firebase DB which is an unstructured database is used in this project. The application was tested using the npm local server and the test results show that the application works as expected, it was deployed using Heroku as a PaaS (Platform as a Service) so as to allow the application to be available over the internet.