A UTILIZAÇÃO DA INTELIGÊNCIA ARTIFICIAL NO COMBATE AO PHISHING

Abstract

This paper analyzes the phenomenon of phishing through the lens of Social Engineering, investigating the manipulation of cognitive biases for sensitive data exfiltration. Through a systematic literature review, the role of Artificial Intelligence (AI) in identifying and mitigating cyber threats is examined. The research stratifies the efficacy of Machine Learning (ML) algorithms, such as Support Vector Machine (SVM) and XGBoost, in statistical analysis and pattern recognition within metadata. Additionally, it explores the application of Long Short-Term Memory (LSTM) networks and Natural Language Processing (NLP) within the Deep Learning (DL) domain, aiming to detect semantic triggers of urgency and coercion in electronic communications. The results indicate that the hybrid integration of ML and DL models enhances predictive accuracy, overcoming the limitations of isolated approaches. Furthermore, the dual nature of AI is discussed, highlighting its ambivalent use: as a robust defense vector and as an automation tool for large-scale spear-phishing attacks. It is concluded that the implementation of AI-based multi-layered systems is imperative for the cyber resilience of users across diverse levels of technical proficiency.