Attackers are constantly developing new strategies to fool users into disclosing critical information, making phishing attacks one of cybersecurity’s most prevalent and severe dangers. Researchers have suggested various techniques for identifying and avoiding phishing attempts, including machine learning-based strategies. This article uses machine learning techniques to study phishing URLs thoroughly. With a dataset of almost 50,00,000 URLs, including both original and phishing URLs, we could extract various features, such as textual, visual, and domain-based features. To enhance the quality of our dataset, we also used several preprocessing techniques, such as feature scaling and URL normalization. Our results provide insights into the most important features for detecting phishing URLs and highlight the importance of using a combination of different. In the end, we assess how resilient our model is to adversarial attacks like URL obfuscation and evasion strategies. We demonstrate that our model is resilient even when subjected to such attacks and can still successfully identify phishing URLs. Overall, our work demonstrates the potential of machine learning-based approaches for enhancing cybersecurity and mitigating the risk of phishing attacks.