Project Report on Machine learning-based phishing detection from URLs Abir Mahmud

Abstract

In these times, there has been a significant explosion of internet-connected devices, reaching from smartphones and IoT devices to cloud networks. Particularly, cybercriminals have increasingly turned their attention to these devices, employing phishing attacks that mark human weaknesses relatively exploiting organization weaknesses. So, a phishing attack, unsuspicious online users are cheated by apparently dependable entities into relating their personal data, such as login identifications and credit card details. This pilfered information becomes a valuable resource for scoring more sophisticated cyberattacks. Although several researchers have future machine learning-based explanations to fight phishing attacks, these approaches often rely on a wide array of features, demanding large computational resources. This renders them impractical for devices with limited processing power. To tackle this challenge, the authors have developed a phishing detection method that successfully identifies phishing attacks using just nine lexical features. It showed tests using the ISCXURL-2016 dataset, covering 11,964 examples of genuine and phishing URLs. Their method was tested with various machine learning classifiers, achieving a remarkable accuracy rate of 99.57% when using the Random Forest algorithm.