A bstractThe collaborative filtering (CF)is one ofthe m ostsuccessfultechnique to predictthe Q uality ofService (Q oS)in Internet of Things (IoT)service recom m endation system s.In w hich,location aw are CF (LA CF)has w idely been paid attention as it takes geographicalposition into consideration.H ow ever,the openness of CF w eb service recom m endation m akes them vulnerable to the injection ofattack profiles consisted by m alicious Q oS values (also identified as shilling attacks). Ifattackers construct the profiles w ith the inform ation oflocation,the LA CF m ay be affected m ore severely than CF. Therefore,to dem onstrate the vulnerability ofLACF to location aw are shilling attack,this paper firstconstructs three kinds ofattack m odels including LA A ,LA B ,and LA R (location aw are -average,bandw agon,and random )attack m odels according to the differentbehaviors ofattackers.Furtherm ore,the paper com pares the im pactofthe classicalshilling attacks and location aw are shilling attacks on LA CF.The experim entalresults on W SD R EA M dataset show the LA CF indeed suffers from shilling attacks and location aw are shilling attacks. B esides, the results show an interesting phenom ena:in com parison w ith classicalshilling attacks,the location aw are shilling attacks do notalw ays m ake m ore influence on the LA CF. K eyw ords: Collaborative filtering,Location aw are shilling attacks,Q oS,W eb service __________________________________________________________________________________________________________________