Trust-based　access　control　(TBAC)　is　a　hot　research　issue　in　the　area　of　network　security　for　open　networks.　Clustering　of　the　domains　of　recommendation　entities　is　a　prerequisite　for　trust　quantification　and　evaluation　in　TBAC　during　the　interaction　of　network　entities.　In　this　paper,　we　propose　a　clustering　algorithm　using　the　membership　function　based　on　rough　sets　for　recommendation　domain　in　TBAC　in　which　traditional　K-mode,　k-mean　and　FCM　clustering　techniques　are　utilized.　First,　based　on　a　system　threshold　for　the　number　of　recommendation　entities,　the　proposed　algorithm　derives　the　clustering　set　for　the　recommendation　entities　within　a　most　recent　minimum　time　window.　Then,　the　domain　for　the　recommendation　entity　of　the　subject　is　obtained　by　analyzing　the　clustering　results　in　which　the　entity　data　samples　that　the　object　gives　to　a　request　is　fixed　as　the　initial　clustering　centers　in　order　to　ensure　the　accuracy　of　the　recommendation　domain　thus　formed.　To　evaluate　the　effectiveness　of　the　proposed　algorithm,　we　apply　both　the　IRIS　standard　data　sets　and　the　UCI　data　sets　in　our　experiment.　And　our　evaluation　results　show　that　the　proposed　algorithm　performs　better　in　terms　of　efficiency　and　clustering　accuracy　compared　to　traditional　K-mode,　K-mean　and　FCM　clustering　algorithms　for　recommendation　domain　in　TBAC.　©　Springer　International　Publishing　Switzerland　2015.