The　open　and　anonymous　nature　of　a　P2P　network　makes　an　ideal　medium　for　many　applications.　The　peers　are　not　willing　to　be　responsible　for　their　historical　behavior　because　of　lack　of　supervisor;　and　it＇s　not　easy　to　set　up　a　trust　mechanism　in　a　traditional　way　to　solve　the　problem　during　the　transaction.　In　this　paper,　we　proposed　a　recommendation-based　trust　model,　and　introduce　some　feathers　to　implement　the　trust　in　P2P　networks.　A　peer＇s　trust　value　is　calculated　according　to　the　interacting　experience　and　recommendation　obtained　from　other　peers.　Compared　with　some　existing　models,　our　approach　can　stimulate　the　feedback　and　decrease　the　dishonest　behaviors　efficiently.　The　computer　simulation　results　show　that　the　trust　model　can　prevent　the　attack　of　malicious　peers＇　recommendation.　Meanwhile,　it　solves　new　peer＇s　trust　problem　and　strengthen　network　security.