Cargando…

De-Identification Mechanism of User Data in Video Systems According to Risk Level for Preventing Leakage of Personal Healthcare Information

A problem with biometric information is that it is more sensitive to external leakage, because it is information that cannot be changed immediately compared to general authentication methods. Regarding facial information, a case in which authentication was permitted by facial information output by a...

Descripción completa

Detalles Bibliográficos
Autores principales: Kim, Jinsu, Park, Namje
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9002924/
https://www.ncbi.nlm.nih.gov/pubmed/35408203
http://dx.doi.org/10.3390/s22072589
Descripción
Sumario:A problem with biometric information is that it is more sensitive to external leakage, because it is information that cannot be changed immediately compared to general authentication methods. Regarding facial information, a case in which authentication was permitted by facial information output by a 3D printer was found. Therefore, a method for minimizing the leakage of biometric information to the outside is required. In this paper, different levels of identification information according to the authority of the user are provided by the de-identification of metadata and face information in stages. For face information and metadata, the level of de-identification is determined and achieved according to the risk level of the de-identified subject. Then, we propose a mechanism to minimize the leakage path by preventing reckless data access by classifying access rights to unidentified data according to four roles. The proposed mechanism provides only differentially de-identified data according to the authority of the accessor, and the required time to perform the de-identification of one image was, on average, 3.6 ms for 300 datapoints, 3.5 ms for 500 datapoints, and 3.47 ms for 1000 datapoints. This confirmed that the required execution time was shortened in proportion to the increase in the size of the dataset. The results for the metadata were similar, and it was confirmed that it took 4.3 ms for 300 cases, 3.78 ms for 500 cases, and 3.5 ms for 1000 cases.