With　the　large-scale　application　of　personalized　recommendation　system,　the　importance　of　user　privacy　protection　has　become　increasingly　prominent.　Among　them,　collaborative　filtering　recommendation　algorithm,　as　one　of　the　widely　used　recommendation　technologies,　not　only　realizes　personalized　recommendation,　but　also　draws　attention　to　its　potential　privacy　disclosure　risk.　Therefore,　how　to　adopt　effective　privacy　protection　policies　within　the　framework　of　collaborative　filtering　to　ensure　the　security　of　user　data　has　become　one　of　the　core　areas　of　current　research.　Existing　research　generally　adopts　the　method　of　adding　noise　to　user-item　rating　data　to　protect　user　privacy,　but　this　usually　leads　to　a　significant　decline　in　recommendation　accuracy.　In　order　to　overcome　the　problem　that　excessive　noise　may　affect　the　quality　of　recommendation　and　even　lead　to　the　failure　of　recommendation,　an　innovative　adaptive　noise　adding　method　is　proposed　in　this　paper.　According　to　the　similarity　difference　between　users,　this　method　specifically　integrates　different　degrees　of　Laplacian　noise　into　the　user　similarity　matrix,　aiming　to　accurately　regulate　the　amount　of　noise　injection,　so　as　to　minimize　the　interference　to　the　original　data　while　maintaining　high　recommendation　accuracy.　In　order　to　further　improve　the　efficiency　of　the　algorithm,　the　algorithm　also　combines　the　optimized　k-means　algorithm　to　construct　a　more　refined　neighbor　set　by　intelligent　clustering　of　users.　The　experimental　results　show　that　the　proposed　algorithm　has　good　performance　and　is　compared　with　other　privacy-protecting　recommendation　algorithms.　The　experimental　results　confirm　the　effectiveness　and　feasibility　of　the　algorithm,　indicating　that　it　can　maintain　high　quality　personalized　recommendation　effect　while　protecting　user　privacy.　©　2024　IEEE.