Machine-Learning-Driven Detection of Malicious Emails and Fake Websites

This research introduces an innovative way to combat phishing attacks and malicious emails by using an ML-driven system. It adopts a unique comprehensive approach, which encompasses data set collection, pre-processing, ML, model, development, and user-friendly browser, and extension interface. We have achieved a striking 96.4% accuracy in identifying Phishing URLs through XGBoost and 98.6% accuracy in detecting malicious email content using Multinomial Naive Bayes, which exemplifies heightened cybersecurity protocols. The dataset mirrors diverse phishing scenarios, guided by Exploratory Data Analysis (EDA). The URL features are utilized by XGBoost model for precise phishing website identification. Multinomial Naive Bayes is found to excel in discerning content of malicious emails. The implementation has been done as a browser extension which ensures seamless integration into user’s digital landscapes, which ensures immediate feedback on the threats encountered for swift mitigation and user awareness.