AI를 활용한 해킹메일 대응 훈련의 실효성에 대한 연구

초록

This study empirically examines whether AI-based phishing response training provides more effective learning experiences than conventional manual training, using large-scale data from 12,180 public institution employees, and proposes an automated training framework. Existing manual training has structural limitations in both content diversity and measurement metrics. Recent studies demonstrate that large language models (LLMs) can automatically generate expert-level persuasive text, and since phishing emails are fundamentally persuasive communication, this capability can be leveraged as a training tool. Results show that participants exposed to AI-based training perceived threats more realistically and demonstrated statistically significant improvements in training effectiveness perception and threat identification capability. AI-generated scenarios combining real-time information and personalization maximized training realism through complex social engineering elements, consistent with prior research on LLM persuasion. Based on these findings, this study proposes the Intelligent & Adaptive Phishing Training Framework (IATF) reflecting the core principles of NIST SP 800-50r1.

키워드