GraphRAG 기술을 활용한 산업기술보호 컴플라이언스 적용 연구

As the competition for technological supremacy intensifies and industrial technology emerges as acore asset for national security, the government is strengthening its protection system through the “Acton the Prevention of Leakage and Protection of Industrial Technology.” However, the complexhierarchical structure and cross-referential nature of the legislation create limitations for practitionersin fully understanding and complying with all regulations, leading to technology leakage due tointernal oversight and human error. Therefore, this study proposes a security compliance automatedevaluation model based on GraphRAG, utilizing a specialized Knowledge Graph built for the IndustrialTechnology Protection Act domain. The proposed model structures unstructured text such as laws,enforcement decrees, and guidelines into nodes and edges, then performs keyword searches and graphtraversal to extract the legal basis essential for evaluation. Performance evaluation achieved a highAdherence score of 0.9688 and an Accuracy of 0.7661. This indicates the model makes reliablejudgments based on retrieved legal grounds without hallucination. By presenting specific reasons andgrounds for violations, it helps prevent the risk of unintentional technology leakage in advance.