In　this　paper,　we　propose　a　permission-based　malware　detection　framework　for　Android　platform.　The　proposed　framework　uses　PCA(Principal　Component　Analysis)　algorithm　for　features　selection　after　permissions　extracted,　and　applies　SVM(support　vector　machine)　methods　to　classify　the　collected　data　as　benign　or　malicious　in　the　process　of　detection.　The　simulation　experimental　results　suggest　that　this　proposed　detection　framework　is　effective　in　detecting　unknown　malware,　and　compared　with　traditional　antivirus　software,　it　can　detect　unknown　malware　effectively　and　immediately　without　updating　the　newest　malware　sample　library　in　time.　It　also　illustrates　that　using　permissions　features　alone　with　machine　learning　methods　can　achieve　good　detection　result.