Algorithms Based on Artificial Intelligence for the Detection and Prevention of Social Engineering Attacks: Systematic review

Abstract

In this study, the growing challenge of cybersecurity is addressed by reviewing Artificial Intelligence (AI) based algorithms designed for the detection and prevention of Social Engineering attacks. The research focuses on identifying effective algorithms, with special attention to phishing, a widely prevalent type of attack. Using the PICOC framework, initially, 891 articles from SCOPUS were collected, of which, after applying rigorous criteria through the Prisma methodology, 32 were selected for detailed analysis. The results reveal that among the studied algorithms, XGBoost, Random Forest (RF), and the combination of FastText-CBOW with Random Forest stand out, exhibiting accuracy rates exceeding 99% in the detection of social engineering attacks. This analysis supports the effectiveness of AI-based tools compared to traditional methods, especially in situations of immediate or 'Zero Hour' attacks. In conclusion, AI emerges as a significant alternative to strengthen cybersecurity and protect against increasingly sophisticated threats.