Examination of Machine Learning based email security appliances root causes of low efficiency against Social Engineering attacks

Abstract

https://doi.org/10.12700/btsz.2025.7.1.SI.73

I have come across many IT solutions that were not able to react effectively to Social Engineering attacks. However, if we examine the process, a malicious email sent by an attacker could be intercepted in many places, depending on the type. For example, I tried to send a malicious email to a business laptop with average protection and I was surprised to find that no alarm was indicated. In my secondary rese- arch, I got to know the scientific journal articles related to the deep learning-based detection of Social Engineering in the literature, and
after a precise and deep understanding of the technologies, I started to develop a methodology to solve the problem. As my primary research, I conducted interviews about the topicality of the topic and during the interviews I tried to map the current domestic email security situation. Also, I performed measurements on what seemed to me to be suspiciously high precision models in order to make sure of the accuracy of the results.