A Comparative Study for Optimizing Phishing Detection Using Machine Learning Algorithms

Phishing is a deceptive practice where cybercriminals disguise themselves as trustworthy entities to acquire sensitive information and pose a severe threat to individuals and organizations. This deceptive technique has evolved and has become increasingly sophisticated, targeting unsuspecting users through various channels, including emails, messages, and fraudulent websites. The term "phishing" first appeared in the mid-1990s, using software tools such as AOHell to steal AOL user names and passwords. Since then, Phishing has become a widespread and constantly changing threat, demanding continuous attention and adaptable cybersecurity measures to prevent malicious actors and safeguard digital assets. In this paper, we developed a model for detecting phishing attacks. We used a variety of ML algorithms, including random forest (RF), decision tree (DT), MLP, and SVM. We used a standard dataset of legitimate phishing attacks obtained from Kaggle to assess the model’s performance, which contained 11,000 observations and 31 distinct features. The random forest algorithm performed admirably, with a maximum accuracy of 97% across the 31 features