Random Forest Algorithm for Email Threat Detection

Abstract

Cybercrime manipulation is on the rise, especially email-based threats; traditional security systems are struggling to keep up with the challenges posed by these activities and secure the cyber world. The paper presents a Random Forest algorithm for detecting and preventing malicious email activity. To bridge this gap, the study used a combination of rule-based filtering and machine-learning classification to analyse a dataset of over 10,000 emails, using Python for system deployment and Random Forest for classification, with some flagged as malicious. The Random Forest model was trained and tested using a 60/40 data split and showed impressive results, with an accuracy of 88.04%, an AUC of 0.973, and an F1-score of 0.879, outperforming other model classifiers such as decision trees and confusion matrices. The study suggests that machine learning, particularly Random Forest, can significantly improve email threat detection and help protect cyberspace. These affirm the potential of machine learning, particularly Random Forest, in enhancing email threat detection and forecasting potential victims.