摘要: To ease the burden of implementing and maintaining access-control aspects in a system, growing trend among developers is to write policies specification language such as XACML integrate with applications through use policy decision point (PDP). assure that specified polices reflect expected ones, recent research has developed verification tools; however, their practice are still limited, being constrained by limited set supported features unavailability properties. This paper presents data-mining approach problem verifying expressed true desires author. We tool investigate this automatically generating requests, evaluating those requests get responses, applying machine learning on request-response pairs infer These inferred properties facilitate inspection behavior. applied our an central grades repository system for university. Our results show algorithms can provide valuable insight into basic help identify specific bug-exposing requests.