Enhanced machine learning algorithm for detection and classification of phishing attacks

Abstract

Machine learning employs artificial neural networks to acquire representations. Phishing is duplicitous behavior or threat where attackers aim to obtain credential information from websites. Phishing websites are fraudulent attempts by cybercriminals to impersonate reputable sites with mission of deceiving victims in divulging personnel information such as credentials, and personal data. Detecting and categorizing these malicious sites has been a topic of interest, with various methods focusing on URL-based techniques proving effective. Among these, machine learning and artificial intelligence approaches leveraging URL features have demonstrated superior results, contingent on the specific features utilized. This research proposes a novel model using a decision tree and a specific feature set to enhance the accuracy of phishing website detection, especially IoT devices. The research explores the impact of selecting a the most common attributes from the well-trained datasets to optimize performance and speed in classification and categorization phishing attacks on IoT devices. Experimental findings and comparative analysis present that the implemented algorithms achieve exceptional performance, with the proposed model achieving an impressive accuracy in identifying phishing URLs.