Abstract: Results of role mining in largescale organizations could not reflect requirements of system security in existing approaches to bottom-up role engineering,and system administration tasks were very arduous.In order to optimize role mining,an optimization method was proposed from the enduser perspective,which converted the role mining into clustering problem.Useroriented roles constraints and four-tuples of clustering were utilized to optimize the mining process and evaluate the mining effect comprehensively.It was tested on several datasets,both synthetic and real.Experimental results showed that,the method not only reduces the burden of administration, but also ensures system security.