Due　to　the　emergence　of　more　and　more　untrustworthy　cloud　service　providers.　How　to　avoid　these　untrustworthy　cloud　services　when　choosing　cloud　services　has　become　an　urgent　concern　of　researchers.　QoS　as　a　criterion　for　evaluating　a　quality　of　service　is　often　used　as　a　basis　for　selection.　Collaborative　Filtering　is　a　common　algorithm　to　predict　QoS　in　existing　research.　However,　the　traditional　collaborative　filtering　covers　some　untrusted　data　in　the　prediction,　which　will　bring　deviation　to　the　prediction　results.　In　this　paper,　sliding　window　is　introduced　on　the　basis　of　the　collaborative　filtering　to　deal　with　untrusted　data　to　improve　the　accuracy　of　prediction　©　2022　IEEE.