메타패스 기반 데이터 비식별화 기술 특허 분석

With the growing scale of data collection and utilization, data de-identification technologies have emerged as essential components across industrial and public sectors. In particular, the expansion of artificial intelligence and data-driven services has accelerated the number of patent applications related to de-identification technologies. This study constructs a heterogeneous patent graph incorporating patents, technology classification codes, applicants, and technical keywords by using registered patent data from the previous three years, and quantitatively analyzes the structural relationships among de-identification technologies through meta-path-based analysis. PathSim similarities were computed for each meta path, and the Louvain algorithm was applied to derive representative keyword communities, revealing that de-identification technologies exhibit distinct organizational structures depending on whether they are viewed from the perspective of classification schemes or industrial actors. The findings provide a quantitative analytical framework for understanding the multilayered relational structure of the de-identification technology patent landscape, and they offer foundational insights for guiding future research and industrial applications in privacy-enhancing technologies.