Phishing is a prevalent cyber-attack method where attackers disguise themselves as trustworthy entities to deceive individuals into divulging sensitive information such as usernames, passwords, and financial details. This study aims to enhance phishing detection and prevention mechanisms by identifying the most significant features that distinguish phishing websites from legitimate ones and by evaluating the effectiveness of different machine learning models in phishing detection. A quantitative data analysis approach was employed, leveraging a dataset from the UCI Machine Learning Repository. The study focused on attributes such as SSL/TLS certificates, URL manipulation, and subdomain usage to identify potential phishing sites. The Random Forest model and Logistic Regression model were compared to determine their accuracy and reliability in detecting phishing websites. The results indicate that the Random Forest model outperforms the Logistic Regression model in terms of precision, recall, F1 score, and accuracy, making it a more robust tool for phishing detection. This research highlights the need for continuous updating of detection models and datasets to keep pace with the evolving tactics of phishing attacks, ultimately contributing to the development of more effective antiphishing strategies.