Next Generation of Phishing Attacks using AI powered Browsers

TL;DR

This paper presents a real-time browser extension utilizing a machine learning model, specifically Random Forest, achieving high accuracy in detecting phishing websites, including zero-day threats, outperforming existing security measures.

Contribution

It introduces a novel real-time browser extension with a machine learning model that effectively detects phishing websites, including previously unseen attacks, enhancing cybersecurity defenses.

Findings

Random Forest achieved 98.32% accuracy in phishing detection.

The model detected zero-day phishing attacks with 99.11% accuracy.

Outperformed Google Safe Browsing in detecting evasive phishing URLs.

Abstract

The increase in the number of phishing demands innovative solutions to safeguard users from phishing attacks. This study explores the development and utilization of a real-time browser extension integrated with machine learning model to improve the detection of phishing websites. The results showed that the model had an accuracy of 98.32%, precision of 98.62%, recall of 97.86%, and an F1-score of 98.24%. When compared to other algorithms like Support Vector Machine, Na\"ive Bayes, Decision Tree, XGBoost, and K Nearest Neighbor, the Random Forest algorithm stood out for its effectiveness in detecting phishing attacks. The zero-day phishing attack detection testing over a 15-day period revealed the model's capability to identify previously unseen threats and thus achieving an overall accuracy rate of 99.11%. Furthermore, the model showed better performance when compared to conventional security measures like Google Safe Browsing. The model had successfully detected phishing URLs that evaded detection by Google safe browsing. This research shows how using machine learning in real-time browser extensions can defend against phishing attacks. It gives useful information about cybersecurity and helps make the internet safer for everyone.