Developing a Phishing Detection System Utilizing a Hybrid Machine Learning Approach Focused on URL Analysis

Cybercrimes of all kinds are coordinated these days via the internet. Thus, phishing attempts are the main focus of this research. The main technique used in phishing attacks is email distortion. Mock sites are used in conjunction with challenging correspondences to collect the requisite data from the relevant parties. Despite the fact that several research on the prevention, detection, and awareness of phishing assaults have been published, there is still no comprehensive and reliable way for doing so. Machine learning is therefore crucial to the battle against cybercrimes such as phishing. The phishing URL-based dataset, a compilation of phishing and authentic URL attributes gathered from over 11,000 domain datasets, serves as the foundation for the proposed study. After preprocessing, a number of machine learning techniques have been applied to guard against phishing URLs and safeguard users. In order to effectively and accurately defend against phishing attacks, this study makes use of a variety of machine learning models, including decision trees, logistic regression, random forests, naive Bayes, gradient boosting classifiers, K-neighbors' classifiers, and support vector classifiers. Additionally, a hybrid LSD model that combines decision trees, support vector machines, and logistic regression with both soft and hard voting is proposed. The grid searches hyper parameter optimization methodology and the canopy feature selection method with cross-fold validation are used in the suggested LSD model. To further illustrate the impacts and efficacy of the models, an array of evaluation criteria was applied to evaluate the suggested technique. Precision, accuracy, recall, F1-score, and specificity were among these requirements. The results of the comparative analysis demonstrate that the suggested approach outperforms the alternative models and gives the best results