The Machine Learning Techniques to Detect Social Engineering Attacks in Text-Based Communications: A Systematic Review

Abstract

In the digital age, social engineering-type attacks have posed a significant threat in the cybersecurity space. The act of Social Engineering attacks is the act of leveraging psychological manipulation to deceive individuals into divulging confidential information or performing harmful actions. Detecting Social Engineering is challenging due to the emergence of artificial intelligence and contextual subtleties. Therefore, to improve on cybersecurity posture, this systematic review explores (ML) machine learning techniques designed to identify social engineering attacks in text-based communications, by analysing the performance, methodologies, and limitations of ML techniques. Machine learning techniques in the detection of social engineering attacks have progressed from simple lexical classifiers to sophisticated deep contextual models. The engine of this paper is an empirical systematic review that identifies trends, gaps, and strengths in current literature. Paucity speaking on literature, PRIMSA is used for systematically surveying articles aligned with the detection of social engineering attacks using ML techniques. The ability of machine learning techniques to effectively identify various forms of SE attacks and adapt to emerging threats makes ML a great tool in combating Social Engineering attacks. As the volume of digital communication persistently grows at an unprecedented rate, so does the potential of criminals to exploit these channels. The researchers concluded with a recommendation for a comprehensive survey of ML techniques for detecting social engineering attacks in text-based communications.