Detecting Cyber Threats Utilizing Machine Learning Approaches: An Assessment of Performance Perspective

Abstract

In contemporary society, our extensive dependence on the internet for various facets of everyday life has led to a remarkable upswing in online activities. Nevertheless, this surge in internet usage has concurrently resulted in a heightened prevalence of cyber threats and cybercrimes. Cybercriminals persistently devise methods to evade security protocols, rendering conventional approaches insufficient for identifying attacks, particularly those exploiting undisclosed vulnerabilities. To confront this issue, a plethora of machine learning techniques has been devised to fortify cybersecurity and uncover instances of cybercrimes. This study specifically centers on the assessment of three widely adopted machine learning methodologies: Belief Networks, Decision Trees, and Support Vector Machines. Their efficacy in discerning spam messages, detecting intrusions into computer systems, and identifying malicious software is evaluated using established datasets commonly utilized for benchmarking purposes.