Vladimir Marevic

Phishing is a form of social engineering used to trick people into disclosing sensitive or confidential information, which attackers can exploit for financial gain or more advanced cyber threats, such as ransomware attacks. Cybercriminals often impersonate reputable entities to manipulate users into taking harmful actions. These attacks can be broadly targeted or more specialized, such as spear phishing (targeting specific individuals) and whaling (targeting high-ranking executives). With the rise of generative artificial intelligence (AI), particularly conversational chatbots, the ability to craft phishing messages has significantly improved. Attackers can now leverage tools like large language models (LLMs) to generate phishing messages with greater efficiency, fewer grammatical errors, and enhanced personalization, making them more convincing and harder to detect. This development presents new challenges for cybersecurity and underscores the need for advanced defense mechanisms against AI-generated phishing threats. The goal of this research is to investigate the possibilities of using machine learning (ML) classification methods in conjunction with natural language processing (NLP) for phishing attack identification. Metaheuristic approach is used to manage hyperparameter optimization because the effectiveness of ML classifiers depends on the appropriate selection of hyperparameters. Real-world data simulations yield encouraging findings, with the best models achieving accuracy levels above 96.9%.