INFERRING SECURITY POLICIES FROM SEMANTIC ATTRIBUTES

Techniques for inferring security policies from semantic attributes are provided. In one aspect, a method for building a query component executable by a processor is provided. The method includes the steps of: (a) providing subjects and permissions related to making a security policy decision, as well as a training set of permission-to-subject assignments, as inputs to the security policy query component; (b) extracting semantic attributes from natural language freeform text descriptions of the subjects and the permissions; and (c) using machine learning to build the security policy query component based on the permission-to-subject assignments in the training set and the semantic attributes extracted in step (b).