Towards Enhanced Human Mitigation of Vishing Attacks: Leveraging Large Language Models for Real-Time User Guidance

Vishing attacks, a prevalent manifestation of social engineering, exploit human trust and manipulation over phone calls to illicitly obtain sensitive information. As these attacks evolve in sophistication, traditional defense mechanisms struggle to maintain efficacy, necessitating the exploration of alternative solutions. In this context, Large Language Models (LLMs) emerge as a cornerstone for fortifying defenses against vishing attacks. Through harnessing the profound linguistic knowledge embedded within LLMs, there exists the potential to comprehensively analyze conversations, identify subtle indicators characteristic of vishing, and dynamically generate adaptive countermeasures in real-time. This position paper underscores the promising role of LLMs in enhancing cybersecurity defenses against vishing, thereby laying the groundwork for further exploration and advancement in this critical domain.