The great amount of data managed by Big Data technologies have to be correctly assured in order to protect critical enterprise and personal information. Nevertheless, current security solutions for Big Data technologies such as NoSQL databases do not take into account the special characteristics of these technologies. In this paper, we focus on assuring NoSQL document databases proposing a framework composed of three stages: (1) the source data set is analysed by using Natural Language Processing techniques and ontological resources in order to detect sensitive data. (2) we define a metamodel for document NoSQL databases that allows designer to specify both structural and security aspects. (3) this model is implemented into a specific document database tool, MongoDB. Finally, we apply the framework proposed to a case study with a dataset of medical domain.