Publication

Abstract :

The internet has become an indispensable part of our everyday lives, but it has also made it easy to carry out malicious activities, such as phishing, covertly. Phishers utilize social engineering tactics or create fictitious websites to fool their victims into divulging personal or corporate information such as usernames, passwords, and account IDs. Phishers have devised methods to evade the many measures that have been implemented to detect phishing websites. One of the finest methods for identifying these risky behaviors is machine learning. This is done so that machine learning algorithms can identify the characteristics that most phishing attacks have in common. The main conclusions from this work are to investigate different machine learning models, do exploratory data analysis on phishing datasets, and understand their features. I learned a lot about the variables that influence models’ ability to predict whether a URL is secure or not from creating this notebook. I also gained knowledge on how to modify models and how that affects how well they work. According to the Phishing dataset, certain characteristics—like “HTTPS,” “Anchor URL,” and “Website Traffic” are more crucial than others in identifying if a URL is phishing or not. The probability of harmful attachments is decreased by the Gradient Boosting Classifier, which correctly classifies URLs for 97.4% of their respective classes. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.