The symbiosis of artificial intelligence and psychological manipulation in modern phishing attacks

Abstract

The integration of artificial intelligence into cybercrime has brought new challenges. Phishing attacks are no longer just a technical problem; they have become sophisticated influence operations (Brundage et al., 2018). This thesis explores how AI systems combine technical capabilities with an understanding of human psychology to create more effective attacks. The goal was to analyse the latest techniques that use machine learning algorithms to personalise scams and the psychological principles they exploit. The methodology included a literature review, analysis of documented attacks, and comparison with traditional methods. The results show that AI drastically increases the effectiveness of attacks through hyper-personalisation and convincing simulation of authority. The research shows that Cialdini's principles of influence remain effective even when people are aware of social engineering attempts (Mollazehi et al., 2024). Deepfake technology allows the creation of fake video calls that have led to losses of millions of dollars, as demonstrated by highprofile cases affecting major companies (CNN Business, 2024), such as Arup, which lost $25.6 million to a sophisticated deepfake attack. Most importantly, AIenabled attacks are becoming increasingly sophisticated, requiring the development of new approaches in machine learning technologies for their effective detection (Tamal et al., 2024). Traditional defence methods are not sufficient against these attacks. New approaches are needed that overcome the limitations of traditional training methods, which have been shown to be ineffective in providing long-term protection against attacks (Caputo et al., 2014). The work contributes to the understanding of how technology and psychology together create new threats, but also how they can work together to protect us.