TBAC: A Fine-Grained Topic-Based Access Control Model for Text Data

Insufficient authorization and overauthorization are two main problems to be solved in access control systems. If the authorization is too strict, users might not be able to access data that should be accessible. If the authorization is too lax, users might obtain too many access rights, which may cause considerable risks. Finer-grained access control models are needed to solve these problems. In this paper, aiming at insufficient authorization in text databases, we propose the topic-based access control (TBAC) model and two implementation methods of the model (subject-to-object topic-based access control method PD-TBAC and object-to-subject topic-based access control method FD-TBAC). In the TBAC model, the access control decision for each user against each file is totally content-driven. We use the latent Dirichlet allocation (LDA) algorithm to extract topics from each paragraph in each file, and these topics are used to determine users' access rights. Experimental results show that the access control granularity of TBAC is more than 4 times that of the existing content-based access control model.