Machine Learning Defence: Identification Phishing URLs Before They Strike
Abstract
Phishing attacks are a common and effective form of cyber-attack, which can cause significant harm to individuals, organizations, and society as a whole. Traditional defences against phishing attacks have proven to be insufficient in detecting and preventing these attacks, as attackers constantly find new ways to deceive their victims. Machine learning has emerged as a promising approach to identifying phishing URLs and preventing them from causing harm. This study proposes and evaluates a machine learning-based approach to identifying phishing URLs before they can cause damage. We collected a dataset of over 10,000 URLs, which included both phishing and legitimate URLs, and used various machine learning algorithms, including decision trees, random forests, and support vector machines, to classify URLs. Our approach involved analysing a variety of features of URLs, including domain names, subdomains, and query parameters. We evaluated the performance of our approach using standard metrics and compared it to traditional defences, such as blacklisting. Our results show that our approach is effective in detecting phishing URLs and outperforms traditional defences.
Keywords
Machine Learning, Defense