Machine Learning Approaches for Detecting Phishing Attacks in Cybersecurity

Despite the relevance of cyber security, phishing attacks remain an evolving threat that exploits human vulnerability, leading users to unveil their sensitive information. This paper applies machine learning algorithms to recognize phishing attacks, investigating three standard models: Logistic Regression, Decision Tree, and Random Forest. We trained and tested these models on features from phishing and legitimate websites to find relative performance in identifying phishing sites of the tested algorithms. Logistic Regression was the best performer overall, with maximized accuracy and reliability in detecting phishing attacks. Though Random Forest performed well, Logistic Regression outperformed it on classification accuracy by balancing efficiency and robustness. The study proves the applicability of Logistic Regression as a scalable and efficient solution for phishing detection, which means that its application in real-time use cases with fewer computational resources is feasible. The finding underlines the valuable role that machine learning can play in enriching cybersecurity solutions and contributes to the global effort of digital security infrastructures in their fight against phishing.