Linear Policy Recommender Scheme for Large-Scale Attribute-Based Access Control

In the large-scale data sharing platform, access control mechanism plays an important role in protecting data security and privacy. Significantly, Attribute-Based Access Control (ABAC) can support fine-grained access control, which would make the data sharing platform more flexible, efficient and manageable. However, in ABAC, data owner need to manually assign access policies for each data, which would incur lots of workload and limit the usability of the system. Thus, we propose a linear policy recommender scheme for ABAC in this work. Firstly, we propose a general form of access policy named linear policy, which describes the policy as a linear function. Comparing with other forms of policy, linear policy is more flexible and efficient. Secondly, we propose a matrix factorization based linear policy recommender scheme. The scheme learns a policy matrix and a security threshold vector from access logs and recommends the optimal linear policy for each data by a binary matrix factorization model. Intuitively, the policy matrix and security threshold vector can be viewed as the optimal linear policy of each data, which is helpful for ABAC to improve policy generation and management. Finally, sufficient experiments are given to present the performance of the proposed policy recommender system. The result shows that our policy recommender system is efficient and accurate in calculation.