Phishing still poses a huge cybersecurity problem, deliberately exploiting people’s weaknesses such as gathering their sensitive data, including financial information or logins. In this work, we study our work using proactive security mechanisms based on machine learning methods of phishing URL detection. Dataset used are the decision trees, random forests, neural networks, among machines learning techniques which are tested on their ability to separate phishing from the authentic websites. In regard to the ease of feature selection techniques, we implement maximization of interpretability by information gain and model performance by using principal component analysis. It has been found that ensemble methods (specifically random forests) are able to perform very well in the detection of phishing, as they achieve high accuracy and low false positive rates. These results also emphasize that the effect of feature engineering and dataset augmentation on ML’s ability to be resilient to changing phishing strategies, as emphasized in the paper, open up possibilities of effectiveness of MLs as a proactive defender against phishing attacks with information to future avenues of research as well as pragmatic consequences for players in cybersecurity.