Relaxed Safeness in Datalog-Based Policies

This paper presents a safeness condition that is more liberal than the one commonly imposed on Datalog, based on classifying predicate arguments into input and output arguments, thereby extending the expressiveness of Datalog-based policy languages. It is also shown that the relaxed safeness condition is a powerful tool for adding important features to such languages.